diff --git a/.gitignore b/.gitignore index 177e6a8db..4866f6122 100644 --- a/.gitignore +++ b/.gitignore @@ -47,6 +47,7 @@ build* !build-info.cpp.in !build-info.sh !build.zig +!docs/build.md /libllama.so /llama-* android-ndk-* @@ -98,13 +99,14 @@ examples/server/*.mjs.hpp # Python -__pycache__ -.venv -/Pipfile -dist -poetry.lock +/.venv +__pycache__/ +*/poetry.lock poetry.toml +# Nix +/result + # Test binaries /tests/test-backend-ops /tests/test-double-float diff --git a/CMakeLists.txt b/CMakeLists.txt index e3a0cc369..4f6cd6872 100644 --- a/CMakeLists.txt +++ b/CMakeLists.txt @@ -42,6 +42,10 @@ endif() option(BUILD_SHARED_LIBS "build shared libraries" ${BUILD_SHARED_LIBS_DEFAULT}) +if (WIN32) + add_compile_definitions(_CRT_SECURE_NO_WARNINGS) +endif() + # # option list # @@ -152,7 +156,7 @@ install(FILES ${CMAKE_CURRENT_BINARY_DIR}/llama-config.cmake DESTINATION ${CMAKE_INSTALL_LIBDIR}/cmake/llama) install( - FILES convert-hf-to-gguf.py + FILES convert_hf_to_gguf.py PERMISSIONS OWNER_READ OWNER_WRITE diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md index 991d85e49..9ad516049 100644 --- a/CONTRIBUTING.md +++ b/CONTRIBUTING.md @@ -1,14 +1,24 @@ -# Contributing Guidelines +# Pull requests -## Checklist +- Always squash-merge the PR before merging +- Use the following format for your final commit: ` : (#)`. For example: `utils : fix typo in utils.py (#1234)` +- Test your changes: + - Using the commands in the [`tests`](tests) folder. For instance, running the `./tests/test-backend-ops` command tests different backend implementations of the GGML library + - Execute [the full CI locally on your machine](ci/README.md) before publishing +- If the pull request contains only documentation changes (e.g., updating READMEs, adding new wiki pages), please add `[no ci]` to the commit title. This will skip unnecessary CI checks and help reduce build times +- Please rate the complexity of your PR (i.e. `Review Complexity : Low`, `Review Complexity : Medium`, `Review Complexity : High`). This makes it easier for maintainers to triage the PRs. + - The PR template has a series of review complexity checkboxes `[ ]` that [you can mark as](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/about-task-lists) `[X]` for your conveience -* Make sure your PR follows the [coding guidelines](https://github.com/ggerganov/llama.cpp/blob/master/README.md#coding-guidelines) -* Test your changes using the commands in the [`tests`](tests) folder. For instance, running the `./tests/test-backend-ops` command tests different backend implementations of the GGML library -* Execute [the full CI locally on your machine](ci/README.md) before publishing +# Coding guidelines -## PR formatting +- Avoid adding third-party dependencies, extra files, extra headers, etc. +- Always consider cross-compatibility with other operating systems and architectures +- Avoid fancy looking modern STL constructs, use basic `for` loops, avoid templates, keep it simple +- There are no strict rules for the code style, but try to follow the patterns in the code (indentation, spaces, etc.). Vertical alignment makes things more readable and easier to batch edit +- Clean-up any trailing whitespaces, use 4 spaces for indentation, brackets on the same line, `void * ptr`, `int & a` +- Naming usually optimizes for common prefix (see https://github.com/ggerganov/ggml/pull/302#discussion_r1243240963) +- Tensors store data in row-major order. We refer to dimension 0 as columns, 1 as rows, 2 as matrices +- Matrix multiplication is unconventional: [`C = ggml_mul_mat(ctx, A, B)`](https://github.com/ggerganov/llama.cpp/blob/880e352277fc017df4d5794f0c21c44e1eae2b84/ggml.h#L1058-L1064) means $C^T = A B^T \Leftrightarrow C = B A^T.$ + +![matmul](media/matmul.png) -* Please rate the complexity of your PR (i.e. `Review Complexity : Low`, `Review Complexity : Medium`, `Review Complexity : High`). This makes it easier for maintainers to triage the PRs. - - The PR template has a series of review complexity checkboxes `[ ]` that you can mark as `[X]` for your conveience. Refer to [About task lists](https://docs.github.com/en/get-started/writing-on-github/working-with-advanced-formatting/about-task-lists) for more information. -* If the pull request only contains documentation changes (e.g., updating READMEs, adding new wiki pages), please add `[no ci]` to the commit title. This will skip unnecessary CI checks and help reduce build times. -* When squashing multiple commits on merge, use the following format for your commit title: ` : (#)`. For example: `utils : Fix typo in utils.py (#1234)` diff --git a/README.md b/README.md index 3569b2bbb..800b499e9 100644 --- a/README.md +++ b/README.md @@ -13,7 +13,7 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others) > [!IMPORTANT] [2024 Jun 12] Binaries have been renamed w/ a `llama-` prefix. `main` is now `llama-cli`, `server` is `llama-server`, etc (https://github.com/ggerganov/llama.cpp/pull/7809) -### Recent API changes +## Recent API changes - [2024 Jun 26] The source code and CMake build scripts have been restructured https://github.com/ggerganov/llama.cpp/pull/8006 - [2024 Apr 21] `llama_token_to_piece` can now optionally render special tokens https://github.com/ggerganov/llama.cpp/pull/6807 @@ -24,9 +24,9 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others) - [2024 Mar 4] Embeddings API updated https://github.com/ggerganov/llama.cpp/pull/5796 - [2024 Mar 3] `struct llama_context_params` https://github.com/ggerganov/llama.cpp/pull/5849 -### Hot topics +## Hot topics -- **`convert.py` has been deprecated and moved to `examples/convert-legacy-llama.py`, please use `convert-hf-to-gguf.py`** https://github.com/ggerganov/llama.cpp/pull/7430 +- **`convert.py` has been deprecated and moved to `examples/convert_legacy_llama.py`, please use `convert_hf_to_gguf.py`** https://github.com/ggerganov/llama.cpp/pull/7430 - Initial Flash-Attention support: https://github.com/ggerganov/llama.cpp/pull/5021 - BPE pre-tokenization support has been added: https://github.com/ggerganov/llama.cpp/pull/6920 - MoE memory layout has been updated - reconvert models for `mmap` support and regenerate `imatrix` https://github.com/ggerganov/llama.cpp/pull/6387 @@ -39,37 +39,6 @@ Inference of Meta's [LLaMA](https://arxiv.org/abs/2302.13971) model (and others) ---- -
- Table of Contents -
    -
  1. - Description -
    2. -
  2. - Usage - -
    3. -
  3. Contributing
    4. -
  4. Coding guidelines
    5. -
  5. Docs
    6. -
-
- ## Description The main goal of `llama.cpp` is to enable LLM inference with minimal setup and state-of-the-art performance on a wide @@ -87,14 +56,6 @@ Since its [inception](https://github.com/ggerganov/llama.cpp/issues/33#issuecomm improved significantly thanks to many contributions. It is the main playground for developing new features for the [ggml](https://github.com/ggerganov/ggml) library. -**Supported platforms:** - -- [X] Mac OS -- [X] Linux -- [X] Windows (via CMake) -- [X] Docker -- [X] FreeBSD - **Supported models:** Typically finetunes of the base models below are supported as well. @@ -150,12 +111,6 @@ Typically finetunes of the base models below are supported as well. - [x] [Moondream](https://huggingface.co/vikhyatk/moondream2) - [x] [Bunny](https://github.com/BAAI-DCAI/Bunny) -**HTTP server** - -[llama.cpp web server](./examples/server) is a lightweight [OpenAI API](https://github.com/openai/openai-openapi) compatible HTTP server that can be used to serve local models and easily connect them to existing clients. - -[simplechat](./examples/server/public_simplechat) is a simple chat client, which can be used to chat with the model exposed using above web server (use --path to point to simplechat), from a local web browser. - **Bindings:** - Python: [abetlen/llama-cpp-python](https://github.com/abetlen/llama-cpp-python) @@ -224,9 +179,10 @@ Unless otherwise noted these projects are open-source with permissive licensing: - [Paddler](https://github.com/distantmagic/paddler) - Stateful load balancer custom-tailored for llama.cpp ---- +## Demo -Here is a typical run using LLaMA v2 13B on M2 Ultra: +
+Typical run using LLaMA v2 13B on M2 Ultra ``` $ make -j && ./llama-cli -m models/llama-13b-v2/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:\nStep 1:" -n 400 -e @@ -306,454 +262,85 @@ llama_print_timings: eval time = 24513.59 ms / 399 runs ( 61.44 ms llama_print_timings: total time = 25431.49 ms ``` +
+ +
+Demo of running both LLaMA-7B and whisper.cpp on a single M1 Pro MacBook + And here is another demo of running both LLaMA-7B and [whisper.cpp](https://github.com/ggerganov/whisper.cpp) on a single M1 Pro MacBook: https://user-images.githubusercontent.com/1991296/224442907-7693d4be-acaa-4e01-8b4f-add84093ffff.mp4 +
+ ## Usage Here are the end-to-end binary build and model conversion steps for most supported models. -### Get the Code +### Basic usage + +Firstly, you need to get the binary. There are different methods that you can follow: +- Method 1: Clone this repository and build locally, see [how to build](./docs/build.md) +- Method 2: If you are using MacOS or Linux, you can install llama.cpp via [brew, flox or nix](./docs/install.md) +- Method 3: Use a Docker image, see [documentation for Docker](./docs/docker.md) +- Method 4: Download pre-built binary from [releases](https://github.com/ggerganov/llama.cpp/releases) + +You can run a basic completion using this command: ```bash -git clone https://github.com/ggerganov/llama.cpp -cd llama.cpp +llama-cli -m your_model.gguf -p "I believe the meaning of life is" -n 128 + +# Output: +# I believe the meaning of life is to find your own truth and to live in accordance with it. For me, this means being true to myself and following my passions, even if they don't align with societal expectations. I think that's what I love about yoga – it's not just a physical practice, but a spiritual one too. It's about connecting with yourself, listening to your inner voice, and honoring your own unique journey. ``` -### Build +See [this page](./examples/main/README.md) for a full list of parameters. -In order to build llama.cpp you have four different options. +### Conversation mode -- Using `make`: - - On Linux or MacOS: - - ```bash - make - ``` - - - On Windows: - - 1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases). - 2. Extract `w64devkit` on your pc. - 3. Run `w64devkit.exe`. - 4. Use the `cd` command to reach the `llama.cpp` folder. - 5. From here you can run: - ```bash - make - ``` - - - Notes: - - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `make -j 8` will run 8 jobs in parallel. - - For faster repeated compilation, install [ccache](https://ccache.dev/). - - For debug builds, run `make LLAMA_DEBUG=1` - -- Using `CMake`: - - ```bash - cmake -B build - cmake --build build --config Release - ``` - - **Notes**: - - - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `cmake --build build --config Release -j 8` will run 8 jobs in parallel. - - For faster repeated compilation, install [ccache](https://ccache.dev/). - - For debug builds, there are two cases: - - 1. Single-config generators (e.g. default = `Unix Makefiles`; note that they just ignore the `--config` flag): - - ```bash - cmake -B build -DCMAKE_BUILD_TYPE=Debug - cmake --build build - ``` - - 2. Multi-config generators (`-G` param set to Visual Studio, XCode...): - - ```bash - cmake -B build -G "Xcode" - cmake --build build --config Debug - ``` - -- Using `gmake` (FreeBSD): - - 1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics) - 2. Add your user to **video** group - 3. Install compilation dependencies. - - ```bash - sudo pkg install gmake automake autoconf pkgconf llvm15 openblas - - gmake CC=/usr/local/bin/clang15 CXX=/usr/local/bin/clang++15 -j4 - ``` - -### Homebrew - -On Mac and Linux, the homebrew package manager can be used via -``` -brew install llama.cpp -``` -The formula is automatically updated with new `llama.cpp` releases. More info: https://github.com/ggerganov/llama.cpp/discussions/7668 - -### Nix - -On Mac and Linux, the Nix package manager can be used via -``` -nix profile install nixpkgs#llama-cpp -``` -For flake enabled installs. - -Or -``` -nix-env --file '' --install --attr llama-cpp -``` -For non-flake enabled installs. - -This expression is automatically updated within the [nixpkgs repo](https://github.com/NixOS/nixpkgs/blob/nixos-24.05/pkgs/by-name/ll/llama-cpp/package.nix#L164). - -#### Flox - -On Mac and Linux, Flox can be used to install llama.cpp within a Flox environment via -``` -flox install llama-cpp -``` -Flox follows the nixpkgs build of llama.cpp. - -### Metal Build - -On MacOS, Metal is enabled by default. Using Metal makes the computation run on the GPU. -To disable the Metal build at compile time use the `GGML_NO_METAL=1` flag or the `GGML_METAL=OFF` cmake option. - -When built with Metal support, you can explicitly disable GPU inference with the `--n-gpu-layers|-ngl 0` command-line -argument. - -### BLAS Build - -Building the program with BLAS support may lead to some performance improvements in prompt processing using batch sizes higher than 32 (the default is 512). Support with CPU-only BLAS implementations doesn't affect the normal generation performance. We may see generation performance improvements with GPU-involved BLAS implementations, e.g. cuBLAS, hipBLAS. There are currently several different BLAS implementations available for build and use: - -- #### Accelerate Framework: - - This is only available on Mac PCs and it's enabled by default. You can just build using the normal instructions. - -- #### OpenBLAS: - - This provides BLAS acceleration using only the CPU. Make sure to have OpenBLAS installed on your machine. - - - Using `make`: - - On Linux: - ```bash - make GGML_OPENBLAS=1 - ``` - - - On Windows: - - 1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases). - 2. Download the latest version of [OpenBLAS for Windows](https://github.com/xianyi/OpenBLAS/releases). - 3. Extract `w64devkit` on your pc. - 4. From the OpenBLAS zip that you just downloaded copy `libopenblas.a`, located inside the `lib` folder, inside `w64devkit\x86_64-w64-mingw32\lib`. - 5. From the same OpenBLAS zip copy the content of the `include` folder inside `w64devkit\x86_64-w64-mingw32\include`. - 6. Run `w64devkit.exe`. - 7. Use the `cd` command to reach the `llama.cpp` folder. - 8. From here you can run: - - ```bash - make GGML_OPENBLAS=1 - ``` - - - Using `CMake` on Linux: - - ```bash - cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS - cmake --build build --config Release - ``` - -- #### BLIS - - Check [BLIS.md](docs/BLIS.md) for more information. - -- #### SYCL - SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators. - - llama.cpp based on SYCL is used to **support Intel GPU** (Data Center Max series, Flex series, Arc series, Built-in GPU and iGPU). - - For detailed info, please refer to [llama.cpp for SYCL](README-sycl.md). - -- #### Intel oneMKL - Building through oneAPI compilers will make avx_vnni instruction set available for intel processors that do not support avx512 and avx512_vnni. Please note that this build config **does not support Intel GPU**. For Intel GPU support, please refer to [llama.cpp for SYCL](./README-sycl.md). - - - Using manual oneAPI installation: - By default, `GGML_BLAS_VENDOR` is set to `Generic`, so if you already sourced intel environment script and assign `-DGGML_BLAS=ON` in cmake, the mkl version of Blas will automatically been selected. Otherwise please install oneAPI and follow the below steps: - ```bash - source /opt/intel/oneapi/setvars.sh # You can skip this step if in oneapi-basekit docker image, only required for manual installation - cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=Intel10_64lp -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_NATIVE=ON - cmake --build build --config Release - ``` - - - Using oneAPI docker image: - If you do not want to source the environment vars and install oneAPI manually, you can also build the code using intel docker container: [oneAPI-basekit](https://hub.docker.com/r/intel/oneapi-basekit). Then, you can use the commands given above. - - Check [Optimizing and Running LLaMA2 on Intel® CPU](https://www.intel.com/content/www/us/en/content-details/791610/optimizing-and-running-llama2-on-intel-cpu.html) for more information. - -- #### CUDA - - This provides GPU acceleration using the CUDA cores of your Nvidia GPU. Make sure to have the CUDA toolkit installed. You can download it from your Linux distro's package manager (e.g. `apt install nvidia-cuda-toolkit`) or from here: [CUDA Toolkit](https://developer.nvidia.com/cuda-downloads). - - For Jetson user, if you have Jetson Orin, you can try this: [Offical Support](https://www.jetson-ai-lab.com/tutorial_text-generation.html). If you are using an old model(nano/TX2), need some additional operations before compiling. - - - Using `make`: - ```bash - make GGML_CUDA=1 - ``` - - Using `CMake`: - - ```bash - cmake -B build -DGGML_CUDA=ON - cmake --build build --config Release - ``` - - The environment variable [`CUDA_VISIBLE_DEVICES`](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars) can be used to specify which GPU(s) will be used. The following compilation options are also available to tweak performance: - - | Option | Legal values | Default | Description | - |-------------------------------|------------------------|---------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| - | GGML_CUDA_FORCE_DMMV | Boolean | false | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 6.1/Pascal/GTX 1000 or higher). Does not affect k-quants. | - | GGML_CUDA_DMMV_X | Positive integer >= 32 | 32 | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. | - | GGML_CUDA_MMV_Y | Positive integer | 1 | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. | - | GGML_CUDA_FORCE_MMQ | Boolean | false | Force the use of custom matrix multiplication kernels for quantized models instead of FP16 cuBLAS even if there is no int8 tensor core implementation available (affects V100, RDNA3). MMQ kernels are enabled by default on GPUs with int8 tensor core support. With MMQ force enabled, speed for large batch sizes will be worse but VRAM consumption will be lower. | - | GGML_CUDA_FORCE_CUBLAS | Boolean | false | Force the use of FP16 cuBLAS instead of custom matrix multiplication kernels for quantized models | - | GGML_CUDA_F16 | Boolean | false | If enabled, use half-precision floating point arithmetic for the CUDA dequantization + mul mat vec kernels and for the q4_1 and q5_1 matrix matrix multiplication kernels. Can improve performance on relatively recent GPUs. | - | GGML_CUDA_KQUANTS_ITER | 1 or 2 | 2 | Number of values processed per iteration and per CUDA thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs. | - | GGML_CUDA_PEER_MAX_BATCH_SIZE | Positive integer | 128 | Maximum batch size for which to enable peer access between multiple GPUs. Peer access requires either Linux or NVLink. When using NVLink enabling peer access for larger batch sizes is potentially beneficial. | - | GGML_CUDA_FA_ALL_QUANTS | Boolean | false | Compile support for all KV cache quantization type (combinations) for the FlashAttention CUDA kernels. More fine-grained control over KV cache size but compilation takes much longer. | - -- #### hipBLAS - - This provides BLAS acceleration on HIP-supported AMD GPUs. - Make sure to have ROCm installed. - You can download it from your Linux distro's package manager or from here: [ROCm Quick Start (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html#rocm-install-quick). - - - Using `make`: - ```bash - make GGML_HIPBLAS=1 - ``` - - Using `CMake` for Linux (assuming a gfx1030-compatible AMD GPU): - ```bash - HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -R)" \ - cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \ - && cmake --build build --config Release -- -j 16 - ``` - On Linux it is also possible to use unified memory architecture (UMA) to share main memory between the CPU and integrated GPU by setting `-DGGML_HIP_UMA=ON`. - However, this hurts performance for non-integrated GPUs (but enables working with integrated GPUs). - - Note that if you get the following error: - ``` - clang: error: cannot find ROCm device library; provide its path via '--rocm-path' or '--rocm-device-lib-path', or pass '-nogpulib' to build without ROCm device library - ``` - Try searching for a directory under `HIP_PATH` that contains the file - `oclc_abi_version_400.bc`. Then, add the following to the start of the - command: `HIP_DEVICE_LIB_PATH=`, so something - like: - ```bash - HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \ - HIP_DEVICE_LIB_PATH= \ - cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \ - && cmake --build build -- -j 16 - ``` - - - Using `make` (example for target gfx1030, build with 16 CPU threads): - ```bash - make -j16 GGML_HIPBLAS=1 GGML_HIP_UMA=1 AMDGPU_TARGETS=gfx1030 - ``` - - - Using `CMake` for Windows (using x64 Native Tools Command Prompt for VS, and assuming a gfx1100-compatible AMD GPU): - ```bash - set PATH=%HIP_PATH%\bin;%PATH% - cmake -S . -B build -G Ninja -DAMDGPU_TARGETS=gfx1100 -DGGML_HIPBLAS=ON -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Release - cmake --build build - ``` - Make sure that `AMDGPU_TARGETS` is set to the GPU arch you want to compile for. The above example uses `gfx1100` that corresponds to Radeon RX 7900XTX/XT/GRE. You can find a list of targets [here](https://llvm.org/docs/AMDGPUUsage.html#processors) - Find your gpu version string by matching the most significant version information from `rocminfo | grep gfx | head -1 | awk '{print $2}'` with the list of processors, e.g. `gfx1035` maps to `gfx1030`. - - - The environment variable [`HIP_VISIBLE_DEVICES`](https://rocm.docs.amd.com/en/latest/understand/gpu_isolation.html#hip-visible-devices) can be used to specify which GPU(s) will be used. - If your GPU is not officially supported you can use the environment variable [`HSA_OVERRIDE_GFX_VERSION`] set to a similar GPU, for example 10.3.0 on RDNA2 (e.g. gfx1030, gfx1031, or gfx1035) or 11.0.0 on RDNA3. - The following compilation options are also available to tweak performance (yes, they refer to CUDA, not HIP, because it uses the same code as the cuBLAS version above): - - | Option | Legal values | Default | Description | - |------------------------|------------------------|---------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| - | GGML_CUDA_DMMV_X | Positive integer >= 32 | 32 | Number of values in x direction processed by the HIP dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. | - | GGML_CUDA_MMV_Y | Positive integer | 1 | Block size in y direction for the HIP mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants. | - | GGML_CUDA_KQUANTS_ITER | 1 or 2 | 2 | Number of values processed per iteration and per HIP thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs. | - -- #### Vulkan - - **With docker**: - - You don't need to install Vulkan SDK. It will be installed inside the container. - - ```sh - # Build the image - docker build -t llama-cpp-vulkan -f .devops/llama-cli-vulkan.Dockerfile . - - # Then, use it: - docker run -it --rm -v "$(pwd):/app:Z" --device /dev/dri/renderD128:/dev/dri/renderD128 --device /dev/dri/card1:/dev/dri/card1 llama-cpp-vulkan -m "/app/models/YOUR_MODEL_FILE" -p "Building a website can be done in 10 simple steps:" -n 400 -e -ngl 33 - ``` - - **Without docker**: - - Firstly, you need to make sure you have installed [Vulkan SDK](https://vulkan.lunarg.com/doc/view/latest/linux/getting_started_ubuntu.html) - - For example, on Ubuntu 22.04 (jammy), use the command below: - - ```bash - wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key add - - wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list https://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list - apt update -y - apt-get install -y vulkan-sdk - # To verify the installation, use the command below: - vulkaninfo - ``` - - Alternatively your package manager might be able to provide the appropriate libraries. - For example for Ubuntu 22.04 you can install `libvulkan-dev` instead. - For Fedora 40, you can install `vulkan-devel`, `glslc` and `glslang` packages. - - Then, build llama.cpp using the cmake command below: - - ```bash - cmake -B build -DGGML_VULKAN=1 - cmake --build build --config Release - # Test the output binary (with "-ngl 33" to offload all layers to GPU) - ./bin/llama-cli -m "PATH_TO_MODEL" -p "Hi you how are you" -n 50 -e -ngl 33 -t 4 - - # You should see in the output, ggml_vulkan detected your GPU. For example: - # ggml_vulkan: Using Intel(R) Graphics (ADL GT2) | uma: 1 | fp16: 1 | warp size: 32 - ``` - -### Prepare and Quantize - -> [!NOTE] -> You can use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to quantise your model weights without any setup too. It is synced from `llama.cpp` main every 6 hours. - -To obtain the official LLaMA 2 weights please see the Obtaining and using the Facebook LLaMA 2 model section. There is also a large selection of pre-quantized `gguf` models available on Hugging Face. - -Note: `convert.py` has been moved to `examples/convert-legacy-llama.py` and shouldn't be used for anything other than `Llama/Llama2/Mistral` models and their derivatives. -It does not support LLaMA 3, you can use `convert-hf-to-gguf.py` with LLaMA 3 downloaded from Hugging Face. +If you want a more ChatGPT-like experience, you can run in conversation mode by passing `-cnv` as a parameter: ```bash -# obtain the official LLaMA model weights and place them in ./models -ls ./models -llama-2-7b tokenizer_checklist.chk tokenizer.model -# [Optional] for models using BPE tokenizers -ls ./models - vocab.json -# [Optional] for PyTorch .bin models like Mistral-7B -ls ./models - +llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv -# install Python dependencies -python3 -m pip install -r requirements.txt - -# convert the model to ggml FP16 format -python3 convert-hf-to-gguf.py models/mymodel/ - -# quantize the model to 4-bits (using Q4_K_M method) -./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M - -# update the gguf filetype to current version if older version is now unsupported -./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY +# Output: +# > hi, who are you? +# Hi there! I'm your helpful assistant! I'm an AI-powered chatbot designed to assist and provide information to users like you. I'm here to help answer your questions, provide guidance, and offer support on a wide range of topics. I'm a friendly and knowledgeable AI, and I'm always happy to help with anything you need. What's on your mind, and how can I assist you today? +# +# > what is 1+1? +# Easy peasy! The answer to 1+1 is... 2! ``` -### Run the quantized model +By default, the chat template will be taken from the input model. If you want to use another chat template, pass `--chat-template NAME` as a parameter. See the list of [supported templates](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) ```bash -# start inference on a gguf model -./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -n 128 +./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --chat-template chatml ``` -When running the larger models, make sure you have enough disk space to store all the intermediate files. +You can also use your own template via in-prefix, in-suffix and reverse-prompt parameters: -### Running on Windows with prebuilt binaries - -You will find prebuilt Windows binaries on the release page. - -Simply download and extract the latest zip package of choice: (e.g. `llama-b1380-bin-win-avx2-x64.zip`) - -From the unzipped folder, open a terminal/cmd window here and place a pre-converted `.gguf` model file. Test out the main example like so: - -``` -.\main -m llama-2-7b.Q4_0.gguf -n 128 +```bash +./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --in-prefix 'User: ' --reverse-prompt 'User:' ``` -### Memory/Disk Requirements +### Web server -As the models are currently fully loaded into memory, you will need adequate disk space to save them and sufficient RAM to load them. At the moment, memory and disk requirements are the same. +[llama.cpp web server](./examples/server/README.md) is a lightweight [OpenAI API](https://github.com/openai/openai-openapi) compatible HTTP server that can be used to serve local models and easily connect them to existing clients. -| Model | Original size | Quantized size (Q4_0) | -|------:|--------------:|----------------------:| -| 7B | 13 GB | 3.9 GB | -| 13B | 24 GB | 7.8 GB | -| 30B | 60 GB | 19.5 GB | -| 65B | 120 GB | 38.5 GB | +Example usage: -### Quantization +```bash +./llama-server -m your_model.gguf --port 8080 -Several quantization methods are supported. They differ in the resulting model disk size and inference speed. - -*(outdated)* - -| Model | Measure | F16 | Q4_0 | Q4_1 | Q5_0 | Q5_1 | Q8_0 | -|------:|--------------|-------:|-------:|-------:|-------:|-------:|-------:| -| 7B | perplexity | 5.9066 | 6.1565 | 6.0912 | 5.9862 | 5.9481 | 5.9070 | -| 7B | file size | 13.0G | 3.5G | 3.9G | 4.3G | 4.7G | 6.7G | -| 7B | ms/tok @ 4th | 127 | 55 | 54 | 76 | 83 | 72 | -| 7B | ms/tok @ 8th | 122 | 43 | 45 | 52 | 56 | 67 | -| 7B | bits/weight | 16.0 | 4.5 | 5.0 | 5.5 | 6.0 | 8.5 | -| 13B | perplexity | 5.2543 | 5.3860 | 5.3608 | 5.2856 | 5.2706 | 5.2548 | -| 13B | file size | 25.0G | 6.8G | 7.6G | 8.3G | 9.1G | 13G | -| 13B | ms/tok @ 4th | - | 103 | 105 | 148 | 160 | 131 | -| 13B | ms/tok @ 8th | - | 73 | 82 | 98 | 105 | 128 | -| 13B | bits/weight | 16.0 | 4.5 | 5.0 | 5.5 | 6.0 | 8.5 | - -- [k-quants](https://github.com/ggerganov/llama.cpp/pull/1684) -- recent k-quants improvements and new i-quants - - [#2707](https://github.com/ggerganov/llama.cpp/pull/2707) - - [#2807](https://github.com/ggerganov/llama.cpp/pull/2807) - - [#4773 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4773) - - [#4856 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4856) - - [#4861 - importance matrix](https://github.com/ggerganov/llama.cpp/pull/4861) - - [#4872 - MoE models](https://github.com/ggerganov/llama.cpp/pull/4872) - - [#4897 - 2-bit quantization](https://github.com/ggerganov/llama.cpp/pull/4897) - - [#4930 - imatrix for all k-quants](https://github.com/ggerganov/llama.cpp/pull/4930) - - [#4951 - imatrix on the GPU](https://github.com/ggerganov/llama.cpp/pull/4957) - - [#4969 - imatrix for legacy quants](https://github.com/ggerganov/llama.cpp/pull/4969) - - [#4996 - k-qunats tuning](https://github.com/ggerganov/llama.cpp/pull/4996) - - [#5060 - Q3_K_XS](https://github.com/ggerganov/llama.cpp/pull/5060) - - [#5196 - 3-bit i-quants](https://github.com/ggerganov/llama.cpp/pull/5196) - - [quantization tuning](https://github.com/ggerganov/llama.cpp/pull/5320), [another one](https://github.com/ggerganov/llama.cpp/pull/5334), and [another one](https://github.com/ggerganov/llama.cpp/pull/5361) - -### Perplexity (measuring model quality) - -You can use the `perplexity` example to measure perplexity over a given prompt (lower perplexity is better). -For more information, see [https://huggingface.co/docs/transformers/perplexity](https://huggingface.co/docs/transformers/perplexity). - -The perplexity measurements in table above are done against the `wikitext2` test dataset (https://paperswithcode.com/dataset/wikitext-2), with context length of 512. -The time per token is measured on a MacBook M1 Pro 32GB RAM using 4 and 8 threads. - -#### How to run - -1. Download/extract: https://huggingface.co/datasets/ggml-org/ci/resolve/main/wikitext-2-raw-v1.zip -2. Run `./llama-perplexity -m models/7B/ggml-model-q4_0.gguf -f wiki.test.raw` -3. Output: +# Basic web UI can be accessed via browser: http://localhost:8080 +# Chat completion endpoint: http://localhost:8080/v1/chat/completions ``` -perplexity : calculating perplexity over 655 chunks -24.43 seconds per pass - ETA 4.45 hours -[1]4.5970,[2]5.1807,[3]6.0382,... -``` -And after 4.45 hours, you will have the final perplexity. ### Interactive mode -If you want a more ChatGPT-like experience, you can run in interactive mode by passing `-i` as a parameter. +> [!NOTE] +> If you prefer basic usage, please consider using conversation mode instead of interactive mode + In this mode, you can always interrupt generation by pressing Ctrl+C and entering one or more lines of text, which will be converted into tokens and appended to the current context. You can also specify a *reverse prompt* with the parameter `-r "reverse prompt string"`. This will result in user input being prompted whenever the exact tokens of the reverse prompt string are encountered in the generation. A typical use is to use a prompt that makes LLaMA emulate a chat between multiple users, say Alice and Bob, and pass `-r "Alice:"`. Here is an example of a few-shot interaction, invoked with the command @@ -804,18 +391,70 @@ The `grammars/` folder contains a handful of sample grammars. To write your own, For authoring more complex JSON grammars, you can also check out https://grammar.intrinsiclabs.ai/, a browser app that lets you write TypeScript interfaces which it compiles to GBNF grammars that you can save for local use. Note that the app is built and maintained by members of the community, please file any issues or FRs on [its repo](http://github.com/intrinsiclabsai/gbnfgen) and not this one. -### Obtaining and using the Facebook LLaMA 2 model +## Build -- Refer to [Facebook's LLaMA download page](https://ai.meta.com/resources/models-and-libraries/llama-downloads/) if you want to access the model data. -- Alternatively, if you want to save time and space, you can download already converted and quantized models from [TheBloke](https://huggingface.co/TheBloke), including: - - [LLaMA 2 7B base](https://huggingface.co/TheBloke/Llama-2-7B-GGUF) - - [LLaMA 2 13B base](https://huggingface.co/TheBloke/Llama-2-13B-GGUF) - - [LLaMA 2 70B base](https://huggingface.co/TheBloke/Llama-2-70B-GGUF) - - [LLaMA 2 7B chat](https://huggingface.co/TheBloke/Llama-2-7B-chat-GGUF) - - [LLaMA 2 13B chat](https://huggingface.co/TheBloke/Llama-2-13B-chat-GGUF) - - [LLaMA 2 70B chat](https://huggingface.co/TheBloke/Llama-2-70B-chat-GGUF) +Please refer to [Build llama.cpp locally](./docs/build.md) -### Seminal papers and background on the models +## Supported backends + +| Backend | Target devices | +| --- | --- | +| [Metal](./docs/build.md#metal-build) | Apple Silicon | +| [BLAS](./docs/build.md#blas-build) | All | +| [BLIS](./docs/backend/BLIS.md) | All | +| [SYCL](./docs/backend/SYCL.md) | Intel and Nvidia GPU | +| [CUDA](./docs/build.md#cuda) | Nvidia GPU | +| [hipBLAS](./docs/build.md#hipblas) | AMD GPU | +| [Vulkan](./docs/build.md#vulkan) | GPU | + +## Tools + +### Prepare and Quantize + +> [!NOTE] +> You can use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf-my-repo) space on Hugging Face to quantise your model weights without any setup too. It is synced from `llama.cpp` main every 6 hours. + +To obtain the official LLaMA 2 weights please see the Obtaining and using the Facebook LLaMA 2 model section. There is also a large selection of pre-quantized `gguf` models available on Hugging Face. + +Note: `convert.py` has been moved to `examples/convert_legacy_llama.py` and shouldn't be used for anything other than `Llama/Llama2/Mistral` models and their derivatives. +It does not support LLaMA 3, you can use `convert_hf_to_gguf.py` with LLaMA 3 downloaded from Hugging Face. + +To learn more about quantizing model, [read this documentation](./examples/quantize/README.md) + +### Perplexity (measuring model quality) + +You can use the `perplexity` example to measure perplexity over a given prompt (lower perplexity is better). +For more information, see [https://huggingface.co/docs/transformers/perplexity](https://huggingface.co/docs/transformers/perplexity). + +To learn more how to measure perplexity using llama.cpp, [read this documentation](./examples/perplexity/README.md) + +## Contributing + +- Contributors can open PRs +- Collaborators can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch +- Collaborators will be invited based on contributions +- Any help with managing issues and PRs is very appreciated! +- See [good first issues](https://github.com/ggerganov/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions +- Read the [CONTRIBUTING.md](CONTRIBUTING.md) for more information +- Make sure to read this: [Inference at the edge](https://github.com/ggerganov/llama.cpp/discussions/205) +- A bit of backstory for those who are interested: [Changelog podcast](https://changelog.com/podcast/532) + +## Other documentations + +- [main (cli)](./examples/main/README.md) +- [server](./examples/server/README.md) +- [jeopardy](./examples/jeopardy/README.md) +- [GBNF grammars](./grammars/README.md) + +**Development documentations** + +- [How to build](./docs/build.md) +- [Running on Docker](./docs/docker.md) +- [Build on Android](./docs/android.md) +- [Performance troubleshooting](./docs/token_generation_performance_tips.md) +- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks) + +**Seminal papers and background on the models** If your issue is with model generation quality, then please at least scan the following links and papers to understand the limitations of LLaMA models. This is especially important when choosing an appropriate model size and appreciating both the significant and subtle differences between LLaMA models and ChatGPT: - LLaMA: @@ -826,178 +465,3 @@ If your issue is with model generation quality, then please at least scan the fo - GPT-3.5 / InstructGPT / ChatGPT: - [Aligning language models to follow instructions](https://openai.com/research/instruction-following) - [Training language models to follow instructions with human feedback](https://arxiv.org/abs/2203.02155) - -### Android - -#### Build on Android using Termux -[Termux](https://github.com/termux/termux-app#installation) is a method to execute `llama.cpp` on an Android device (no root required). -``` -apt update && apt upgrade -y -apt install git make cmake -``` - -It's recommended to move your model inside the `~/` directory for best performance: -``` -cd storage/downloads -mv model.gguf ~/ -``` - -[Get the code](https://github.com/ggerganov/llama.cpp#get-the-code) & [follow the Linux build instructions](https://github.com/ggerganov/llama.cpp#build) to build `llama.cpp`. - -#### Building the Project using Android NDK -Obtain the [Android NDK](https://developer.android.com/ndk) and then build with CMake. - -Execute the following commands on your computer to avoid downloading the NDK to your mobile. Alternatively, you can also do this in Termux: -``` -$ mkdir build-android -$ cd build-android -$ export NDK= -$ cmake -DCMAKE_TOOLCHAIN_FILE=$NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS=-march=armv8.4a+dotprod .. -$ make -``` - -Install [termux](https://github.com/termux/termux-app#installation) on your device and run `termux-setup-storage` to get access to your SD card (if Android 11+ then run the command twice). - -Finally, copy these built `llama` binaries and the model file to your device storage. Because the file permissions in the Android sdcard cannot be changed, you can copy the executable files to the `/data/data/com.termux/files/home/bin` path, and then execute the following commands in Termux to add executable permission: - -(Assumed that you have pushed the built executable files to the /sdcard/llama.cpp/bin path using `adb push`) -``` -$cp -r /sdcard/llama.cpp/bin /data/data/com.termux/files/home/ -$cd /data/data/com.termux/files/home/bin -$chmod +x ./* -``` - -Download model [llama-2-7b-chat.Q4_K_M.gguf](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/blob/main/llama-2-7b-chat.Q4_K_M.gguf), and push it to `/sdcard/llama.cpp/`, then move it to `/data/data/com.termux/files/home/model/` - -``` -$mv /sdcard/llama.cpp/llama-2-7b-chat.Q4_K_M.gguf /data/data/com.termux/files/home/model/ -``` - -Now, you can start chatting: -``` -$cd /data/data/com.termux/files/home/bin -$./llama-cli -m ../model/llama-2-7b-chat.Q4_K_M.gguf -n 128 -cml -``` - -Here's a demo of an interactive session running on Pixel 5 phone: - -https://user-images.githubusercontent.com/271616/225014776-1d567049-ad71-4ef2-b050-55b0b3b9274c.mp4 - -### Docker - -#### Prerequisites -* Docker must be installed and running on your system. -* Create a folder to store big models & intermediate files (ex. /llama/models) - -#### Images -We have three Docker images available for this project: - -1. `ghcr.io/ggerganov/llama.cpp:full`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. (platforms: `linux/amd64`, `linux/arm64`) -2. `ghcr.io/ggerganov/llama.cpp:light`: This image only includes the main executable file. (platforms: `linux/amd64`, `linux/arm64`) -3. `ghcr.io/ggerganov/llama.cpp:server`: This image only includes the server executable file. (platforms: `linux/amd64`, `linux/arm64`) - -Additionally, there the following images, similar to the above: - -- `ghcr.io/ggerganov/llama.cpp:full-cuda`: Same as `full` but compiled with CUDA support. (platforms: `linux/amd64`) -- `ghcr.io/ggerganov/llama.cpp:light-cuda`: Same as `light` but compiled with CUDA support. (platforms: `linux/amd64`) -- `ghcr.io/ggerganov/llama.cpp:server-cuda`: Same as `server` but compiled with CUDA support. (platforms: `linux/amd64`) -- `ghcr.io/ggerganov/llama.cpp:full-rocm`: Same as `full` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) -- `ghcr.io/ggerganov/llama.cpp:light-rocm`: Same as `light` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) -- `ghcr.io/ggerganov/llama.cpp:server-rocm`: Same as `server` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) - -The GPU enabled images are not currently tested by CI beyond being built. They are not built with any variation from the ones in the Dockerfiles defined in [.devops/](.devops/) and the GitHub Action defined in [.github/workflows/docker.yml](.github/workflows/docker.yml). If you need different settings (for example, a different CUDA or ROCm library, you'll need to build the images locally for now). - -#### Usage - -The easiest way to download the models, convert them to ggml and optimize them is with the --all-in-one command which includes the full docker image. - -Replace `/path/to/models` below with the actual path where you downloaded the models. - -```bash -docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --all-in-one "/models/" 7B -``` - -On completion, you are ready to play! - -```bash -docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 -``` - -or with a light image: - -```bash -docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 -``` - -or with a server image: - -```bash -docker run -v /path/to/models:/models -p 8000:8000 ghcr.io/ggerganov/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 -``` - -### Docker With CUDA - -Assuming one has the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) properly installed on Linux, or is using a GPU enabled cloud, `cuBLAS` should be accessible inside the container. - -#### Building Locally - -```bash -docker build -t local/llama.cpp:full-cuda -f .devops/full-cuda.Dockerfile . -docker build -t local/llama.cpp:light-cuda -f .devops/llama-cli-cuda.Dockerfile . -docker build -t local/llama.cpp:server-cuda -f .devops/llama-server-cuda.Dockerfile . -``` - -You may want to pass in some different `ARGS`, depending on the CUDA environment supported by your container host, as well as the GPU architecture. - -The defaults are: - -- `CUDA_VERSION` set to `11.7.1` -- `CUDA_DOCKER_ARCH` set to `all` - -The resulting images, are essentially the same as the non-CUDA images: - -1. `local/llama.cpp:full-cuda`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. -2. `local/llama.cpp:light-cuda`: This image only includes the main executable file. -3. `local/llama.cpp:server-cuda`: This image only includes the server executable file. - -#### Usage - -After building locally, Usage is similar to the non-CUDA examples, but you'll need to add the `--gpus` flag. You will also want to use the `--n-gpu-layers` flag. - -```bash -docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1 -docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1 -docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1 -``` - -### Contributing - -- Contributors can open PRs -- Collaborators can push to branches in the `llama.cpp` repo and merge PRs into the `master` branch -- Collaborators will be invited based on contributions -- Any help with managing issues and PRs is very appreciated! -- Make sure to read this: [Inference at the edge](https://github.com/ggerganov/llama.cpp/discussions/205) -- A bit of backstory for those who are interested: [Changelog podcast](https://changelog.com/podcast/532) - -### Coding guidelines - -- Avoid adding third-party dependencies, extra files, extra headers, etc. -- Always consider cross-compatibility with other operating systems and architectures -- Avoid fancy looking modern STL constructs, use basic `for` loops, avoid templates, keep it simple -- There are no strict rules for the code style, but try to follow the patterns in the code (indentation, spaces, etc.). Vertical alignment makes things more readable and easier to batch edit -- Clean-up any trailing whitespaces, use 4 spaces for indentation, brackets on the same line, `void * ptr`, `int & a` -- See [good first issues](https://github.com/ggerganov/llama.cpp/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22) for tasks suitable for first contributions -- Tensors store data in row-major order. We refer to dimension 0 as columns, 1 as rows, 2 as matrices -- Matrix multiplication is unconventional: [`C = ggml_mul_mat(ctx, A, B)`](https://github.com/ggerganov/llama.cpp/blob/880e352277fc017df4d5794f0c21c44e1eae2b84/ggml.h#L1058-L1064) means $C^T = A B^T \Leftrightarrow C = B A^T.$ - -![matmul](media/matmul.png) - -### Docs - -- [main (cli)](./examples/main/README.md) -- [server](./examples/server/README.md) -- [jeopardy](./examples/jeopardy/README.md) -- [BLIS](./docs/BLIS.md) -- [Performance troubleshooting](./docs/token_generation_performance_tips.md) -- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks) -- [GBNF grammars](./grammars/README.md) diff --git a/ci/run.sh b/ci/run.sh index e0cedb24f..9703b77ce 100755 --- a/ci/run.sh +++ b/ci/run.sh @@ -287,7 +287,7 @@ function gg_run_open_llama_7b_v2 { (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} -DGGML_CUDA=1 .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log (time make -j ) 2>&1 | tee -a $OUT/${ci}-make.log - python3 ../examples/convert-legacy-llama.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf + python3 ../examples/convert_legacy_llama.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf model_f16="${path_models}/ggml-model-f16.gguf" model_q8_0="${path_models}/ggml-model-q8_0.gguf" @@ -421,7 +421,7 @@ function gg_run_pythia_1_4b { (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log (time make -j ) 2>&1 | tee -a $OUT/${ci}-make.log - python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf + python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf model_f16="${path_models}/ggml-model-f16.gguf" model_q8_0="${path_models}/ggml-model-q8_0.gguf" @@ -553,7 +553,7 @@ function gg_run_pythia_2_8b { (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} -DGGML_CUDA=1 .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log (time make -j ) 2>&1 | tee -a $OUT/${ci}-make.log - python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf + python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf model_f16="${path_models}/ggml-model-f16.gguf" model_q8_0="${path_models}/ggml-model-q8_0.gguf" @@ -688,7 +688,7 @@ function gg_run_embd_bge_small { (time cmake -DCMAKE_BUILD_TYPE=Release ${CMAKE_EXTRA} .. ) 2>&1 | tee -a $OUT/${ci}-cmake.log (time make -j ) 2>&1 | tee -a $OUT/${ci}-make.log - python3 ../convert-hf-to-gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf + python3 ../convert_hf_to_gguf.py ${path_models} --outfile ${path_models}/ggml-model-f16.gguf model_f16="${path_models}/ggml-model-f16.gguf" model_q8_0="${path_models}/ggml-model-q8_0.gguf" diff --git a/common/common.cpp b/common/common.cpp index 2c05a4d4a..fc0f3b350 100644 --- a/common/common.cpp +++ b/common/common.cpp @@ -190,6 +190,12 @@ int32_t cpu_get_num_math() { // CLI argument parsing // +void gpt_params_handle_hf_token(gpt_params & params) { + if (params.hf_token.empty() && std::getenv("HF_TOKEN")) { + params.hf_token = std::getenv("HF_TOKEN"); + } +} + void gpt_params_handle_model_default(gpt_params & params) { if (!params.hf_repo.empty()) { // short-hand to avoid specifying --hf-file -> default it to --model @@ -237,6 +243,8 @@ bool gpt_params_parse_ex(int argc, char ** argv, gpt_params & params) { gpt_params_handle_model_default(params); + gpt_params_handle_hf_token(params); + if (params.escape) { string_process_escapes(params.prompt); string_process_escapes(params.input_prefix); @@ -472,6 +480,14 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa else { invalid_param = true; } return true; } + if (arg == "--attention") { + CHECK_ARG + std::string value(argv[i]); + /**/ if (value == "causal") { params.attention_type = LLAMA_ATTENTION_TYPE_CAUSAL; } + else if (value == "non-causal") { params.attention_type = LLAMA_ATTENTION_TYPE_NON_CAUSAL; } + else { invalid_param = true; } + return true; + } if (arg == "--defrag-thold" || arg == "-dt") { CHECK_ARG params.defrag_thold = std::stof(argv[i]); @@ -644,6 +660,14 @@ bool gpt_params_find_arg(int argc, char ** argv, const std::string & arg, gpt_pa params.model_url = argv[i]; return true; } + if (arg == "-hft" || arg == "--hf-token") { + if (++i >= argc) { + invalid_param = true; + return true; + } + params.hf_token = argv[i]; + return true; + } if (arg == "-hfr" || arg == "--hf-repo") { CHECK_ARG params.hf_repo = argv[i]; @@ -1394,7 +1418,9 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param options.push_back({ "*", " --keep N", "number of tokens to keep from the initial prompt (default: %d, -1 = all)", params.n_keep }); options.push_back({ "*", " --chunks N", "max number of chunks to process (default: %d, -1 = all)", params.n_chunks }); options.push_back({ "*", "-fa, --flash-attn", "enable Flash Attention (default: %s)", params.flash_attn ? "enabled" : "disabled" }); - options.push_back({ "*", "-p, --prompt PROMPT", "prompt to start generation with (default: '%s')", params.prompt.c_str() }); + options.push_back({ "*", "-p, --prompt PROMPT", "prompt to start generation with\n" + "in conversation mode, this will be used as system prompt\n" + "(default: '%s')", params.prompt.c_str() }); options.push_back({ "*", "-f, --file FNAME", "a file containing the prompt (default: none)" }); options.push_back({ "*", " --in-file FNAME", "an input file (repeat to specify multiple files)" }); options.push_back({ "*", "-bf, --binary-file FNAME", "binary file containing the prompt (default: none)" }); @@ -1409,7 +1435,9 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param "halt generation at PROMPT, return control in interactive mode\n" "can be specified more than once for multiple prompts" }); options.push_back({ "main", "-sp, --special", "special tokens output enabled (default: %s)", params.special ? "true" : "false" }); - options.push_back({ "main", "-cnv, --conversation", "run in conversation mode (does not print special tokens and suffix/prefix, use default chat template) (default: %s)", params.conversation ? "true" : "false" }); + options.push_back({ "main", "-cnv, --conversation", "run in conversation mode, does not print special tokens and suffix/prefix\n" + "if suffix/prefix are not specified, default chat template will be used\n" + "(default: %s)", params.conversation ? "true" : "false" }); options.push_back({ "main infill", "-i, --interactive", "run in interactive mode (default: %s)", params.interactive ? "true" : "false" }); options.push_back({ "main infill", "-if, --interactive-first", "run in interactive mode and wait for input right away (default: %s)", params.interactive_first ? "true" : "false" }); options.push_back({ "main infill", "-mli, --multiline-input", "allows you to write or paste multiple lines without ending each in '\\'" }); @@ -1453,6 +1481,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param options.push_back({ "main", " --cfg-scale N", "strength of guidance (default: %.1f, 1.0 = disable)", (double)sparams.cfg_scale }); options.push_back({ "main", " --chat-template JINJA_TEMPLATE", "set custom jinja chat template (default: template taken from model's metadata)\n" + "if suffix/prefix are specified, template will be disabled\n" "only commonly used templates are accepted:\n" "https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template" }); options.push_back({ "grammar" }); @@ -1463,8 +1492,10 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param "For schemas w/ external $refs, use --grammar + example/json_schema_to_grammar.py instead" }); options.push_back({ "embedding" }); - options.push_back({ "embedding", " --pooling {none,mean,cls}", + options.push_back({ "embedding", " --pooling {none,mean,cls,last}", "pooling type for embeddings, use model default if unspecified" }); + options.push_back({ "embedding", " --attention {causal,non-causal}", + "attention type for embeddings, use model default if unspecified" }); options.push_back({ "context hacking" }); options.push_back({ "*", " --rope-scaling {none,linear,yarn}", @@ -1561,6 +1592,7 @@ void gpt_params_print_usage(int /*argc*/, char ** argv, const gpt_params & param options.push_back({ "*", "-mu, --model-url MODEL_URL", "model download url (default: unused)" }); options.push_back({ "*", "-hfr, --hf-repo REPO", "Hugging Face model repository (default: unused)" }); options.push_back({ "*", "-hff, --hf-file FILE", "Hugging Face model file (default: unused)" }); + options.push_back({ "*", "-hft, --hf-token TOKEN", "Hugging Face access token (default: value from HF_TOKEN environment variable)" }); options.push_back({ "retrieval" }); options.push_back({ "retrieval", " --context-file FNAME", "file to load context from (repeat to specify multiple files)" }); @@ -2000,9 +2032,9 @@ std::tuple llama_init_from_gpt_par llama_model * model = nullptr; if (!params.hf_repo.empty() && !params.hf_file.empty()) { - model = llama_load_model_from_hf(params.hf_repo.c_str(), params.hf_file.c_str(), params.model.c_str(), mparams); + model = llama_load_model_from_hf(params.hf_repo.c_str(), params.hf_file.c_str(), params.model.c_str(), params.hf_token.c_str(), mparams); } else if (!params.model_url.empty()) { - model = llama_load_model_from_url(params.model_url.c_str(), params.model.c_str(), mparams); + model = llama_load_model_from_url(params.model_url.c_str(), params.model.c_str(), params.hf_token.c_str(), mparams); } else { model = llama_load_model_from_file(params.model.c_str(), mparams); } @@ -2070,7 +2102,24 @@ std::tuple llama_init_from_gpt_par if (params.warmup) { LOG("warming up the model with an empty run\n"); - std::vector tmp = { llama_token_bos(model), llama_token_eos(model), }; + std::vector tmp; + llama_token bos = llama_token_bos(model); + llama_token eos = llama_token_eos(model); + // some models (e.g. T5) don't have a BOS token + if (bos != -1) { + tmp.push_back(bos); + } + tmp.push_back(eos); + + if (llama_model_has_encoder(model)) { + llama_encode(lctx, llama_batch_get_one(tmp.data(), tmp.size(), 0, 0)); + llama_token decoder_start_token_id = llama_model_decoder_start_token(model); + if (decoder_start_token_id == -1) { + decoder_start_token_id = bos; + } + tmp.clear(); + tmp.push_back(decoder_start_token_id); + } llama_decode(lctx, llama_batch_get_one(tmp.data(), std::min(tmp.size(), (size_t) params.n_batch), 0, 0)); llama_kv_cache_clear(lctx); llama_synchronize(lctx); @@ -2153,6 +2202,7 @@ struct llama_context_params llama_context_params_from_gpt_params(const gpt_param cparams.yarn_beta_slow = params.yarn_beta_slow; cparams.yarn_orig_ctx = params.yarn_orig_ctx; cparams.pooling_type = params.pooling_type; + cparams.attention_type = params.attention_type; cparams.defrag_thold = params.defrag_thold; cparams.cb_eval = params.cb_eval; cparams.cb_eval_user_data = params.cb_eval_user_data; @@ -2172,7 +2222,7 @@ static bool starts_with(const std::string & str, const std::string & prefix) { return str.rfind(prefix, 0) == 0; } -static bool llama_download_file(const std::string & url, const std::string & path) { +static bool llama_download_file(const std::string & url, const std::string & path, const std::string & hf_token) { // Initialize libcurl std::unique_ptr curl(curl_easy_init(), &curl_easy_cleanup); @@ -2187,6 +2237,15 @@ static bool llama_download_file(const std::string & url, const std::string & pat curl_easy_setopt(curl.get(), CURLOPT_URL, url.c_str()); curl_easy_setopt(curl.get(), CURLOPT_FOLLOWLOCATION, 1L); + // Check if hf-token or bearer-token was specified + if (!hf_token.empty()) { + std::string auth_header = "Authorization: Bearer "; + auth_header += hf_token.c_str(); + struct curl_slist *http_headers = NULL; + http_headers = curl_slist_append(http_headers, auth_header.c_str()); + curl_easy_setopt(curl.get(), CURLOPT_HTTPHEADER, http_headers); + } + #if defined(_WIN32) // CURLSSLOPT_NATIVE_CA tells libcurl to use standard certificate store of // operating system. Currently implemented under MS-Windows. @@ -2382,6 +2441,7 @@ static bool llama_download_file(const std::string & url, const std::string & pat struct llama_model * llama_load_model_from_url( const char * model_url, const char * path_model, + const char * hf_token, const struct llama_model_params & params) { // Basic validation of the model_url if (!model_url || strlen(model_url) == 0) { @@ -2389,7 +2449,7 @@ struct llama_model * llama_load_model_from_url( return NULL; } - if (!llama_download_file(model_url, path_model)) { + if (!llama_download_file(model_url, path_model, hf_token)) { return NULL; } @@ -2437,14 +2497,14 @@ struct llama_model * llama_load_model_from_url( // Prepare download in parallel std::vector> futures_download; for (int idx = 1; idx < n_split; idx++) { - futures_download.push_back(std::async(std::launch::async, [&split_prefix, &split_url_prefix, &n_split](int download_idx) -> bool { + futures_download.push_back(std::async(std::launch::async, [&split_prefix, &split_url_prefix, &n_split, hf_token](int download_idx) -> bool { char split_path[PATH_MAX] = {0}; llama_split_path(split_path, sizeof(split_path), split_prefix, download_idx, n_split); char split_url[LLAMA_CURL_MAX_URL_LENGTH] = {0}; llama_split_path(split_url, sizeof(split_url), split_url_prefix, download_idx, n_split); - return llama_download_file(split_url, split_path); + return llama_download_file(split_url, split_path, hf_token); }, idx)); } @@ -2463,6 +2523,7 @@ struct llama_model * llama_load_model_from_hf( const char * repo, const char * model, const char * path_model, + const char * hf_token, const struct llama_model_params & params) { // construct hugging face model url: // @@ -2478,7 +2539,7 @@ struct llama_model * llama_load_model_from_hf( model_url += "/resolve/main/"; model_url += model; - return llama_load_model_from_url(model_url.c_str(), path_model, params); + return llama_load_model_from_url(model_url.c_str(), path_model, hf_token, params); } #else @@ -2486,6 +2547,7 @@ struct llama_model * llama_load_model_from_hf( struct llama_model * llama_load_model_from_url( const char * /*model_url*/, const char * /*path_model*/, + const char * /*hf_token*/, const struct llama_model_params & /*params*/) { fprintf(stderr, "%s: llama.cpp built without libcurl, downloading from an url not supported.\n", __func__); return nullptr; @@ -2495,6 +2557,7 @@ struct llama_model * llama_load_model_from_hf( const char * /*repo*/, const char * /*model*/, const char * /*path_model*/, + const char * /*hf_token*/, const struct llama_model_params & /*params*/) { fprintf(stderr, "%s: llama.cpp built without libcurl, downloading from Hugging Face not supported.\n", __func__); return nullptr; @@ -2559,51 +2622,35 @@ std::vector llama_tokenize( } std::string llama_token_to_piece(const struct llama_context * ctx, llama_token token, bool special) { - std::vector result(8, 0); - const int n_tokens = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special); - if (n_tokens < 0) { - result.resize(-n_tokens); - int check = llama_token_to_piece(llama_get_model(ctx), token, result.data(), result.size(), special); - GGML_ASSERT(check == -n_tokens); - } else { - result.resize(n_tokens); + std::string piece; + piece.resize(piece.capacity()); // using string internal cache, 15 bytes + '\n' + const int n_chars = llama_token_to_piece(llama_get_model(ctx), token, &piece[0], piece.size(), 0, special); + if (n_chars < 0) { + piece.resize(-n_chars); + int check = llama_token_to_piece(llama_get_model(ctx), token, &piece[0], piece.size(), 0, special); + GGML_ASSERT(check == -n_chars); + } + else { + piece.resize(n_chars); } - return std::string(result.data(), result.size()); + return piece; } -std::string llama_detokenize_spm(llama_context * ctx, const std::vector & tokens) { - const llama_token bos_id = llama_token_bos(llama_get_model(ctx)); - - std::string piece; - std::string result; - - for (size_t i = 0; i < tokens.size(); ++i) { - piece = llama_token_to_piece(ctx, tokens[i]); - - // remove the leading space of the first non-BOS token - if (((tokens[0] == bos_id && i == 1) || (tokens[0] != bos_id && i == 0)) && piece[0] == ' ') { - piece = piece.substr(1); - } - - result += piece; +std::string llama_detokenize(llama_context * ctx, const std::vector & tokens, bool special) { + std::string text; + text.resize(std::max(text.capacity(), tokens.size())); + int32_t n_chars = llama_detokenize(llama_get_model(ctx), tokens.data(), (int32_t)tokens.size(), &text[0], (int32_t)text.size(), false, special); + if (n_chars < 0) { + text.resize(-n_chars); + n_chars = llama_detokenize(llama_get_model(ctx), tokens.data(), (int32_t)tokens.size(), &text[0], (int32_t)text.size(), false, special); + GGML_ASSERT(n_chars <= (int32_t)text.size()); // whitespace trimming is performed after per-token detokenization } - return result; -} - -std::string llama_detokenize_bpe(llama_context * ctx, const std::vector & tokens) { - std::string piece; - std::string result; - - for (size_t i = 0; i < tokens.size(); ++i) { - piece = llama_token_to_piece(ctx, tokens[i]); - - result += piece; - } + text.resize(n_chars); // NOTE: the original tokenizer decodes bytes after collecting the pieces. - return result; + return text; } bool llama_should_add_bos_token(const llama_model * model) { diff --git a/common/common.h b/common/common.h index 65c0ef81a..184a53dc0 100644 --- a/common/common.h +++ b/common/common.h @@ -99,6 +99,7 @@ struct gpt_params { enum llama_split_mode split_mode = LLAMA_SPLIT_MODE_LAYER; // how to split the model across GPUs enum llama_rope_scaling_type rope_scaling_type = LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED; enum llama_pooling_type pooling_type = LLAMA_POOLING_TYPE_UNSPECIFIED; // pooling type for embeddings + enum llama_attention_type attention_type = LLAMA_ATTENTION_TYPE_UNSPECIFIED; // attention type for embeddings // // sampling parameters struct llama_sampling_params sparams; @@ -107,6 +108,7 @@ struct gpt_params { std::string model_draft = ""; // draft model for speculative decoding std::string model_alias = "unknown"; // model alias std::string model_url = ""; // model url to download + std::string hf_token = ""; // HF token std::string hf_repo = ""; // HF repo std::string hf_file = ""; // HF file std::string prompt = ""; @@ -255,6 +257,7 @@ struct gpt_params { bool spm_infill = false; // suffix/prefix/middle pattern for infill }; +void gpt_params_handle_hf_token(gpt_params & params); void gpt_params_handle_model_default(gpt_params & params); bool gpt_params_parse_ex (int argc, char ** argv, gpt_params & params); @@ -310,8 +313,8 @@ std::tuple llama_init_from_gpt_par struct llama_model_params llama_model_params_from_gpt_params (const gpt_params & params); struct llama_context_params llama_context_params_from_gpt_params(const gpt_params & params); -struct llama_model * llama_load_model_from_url(const char * model_url, const char * path_model, const struct llama_model_params & params); -struct llama_model * llama_load_model_from_hf(const char * repo, const char * file, const char * path_model, const struct llama_model_params & params); +struct llama_model * llama_load_model_from_url(const char * model_url, const char * path_model, const char * hf_token, const struct llama_model_params & params); +struct llama_model * llama_load_model_from_hf(const char * repo, const char * file, const char * path_model, const char * hf_token, const struct llama_model_params & params); // Batch utils @@ -349,21 +352,13 @@ std::string llama_token_to_piece( llama_token token, bool special = true); -// TODO: these should be moved in llama.h C-style API under single `llama_detokenize` function -// that takes into account the tokenizer type and decides how to handle the leading space -// // detokenizes a vector of tokens into a string // should work similar to Python's `tokenizer.decode` -// removes the leading space from the first non-BOS token -std::string llama_detokenize_spm( +// optionally renders special/control tokens +std::string llama_detokenize( llama_context * ctx, - const std::vector & tokens); - -// detokenizes a vector of tokens into a string -// should work similar to Python's `tokenizer.decode` -std::string llama_detokenize_bpe( - llama_context * ctx, - const std::vector & tokens); + const std::vector & tokens, + bool special = true); // Uses the value from the model metadata if possible, otherwise // defaults to true when model type is SPM, otherwise false. diff --git a/convert-hf-to-gguf.py b/convert_hf_to_gguf.py similarity index 94% rename from convert-hf-to-gguf.py rename to convert_hf_to_gguf.py index 1ae7abbaf..ffa66cb6d 100755 --- a/convert-hf-to-gguf.py +++ b/convert_hf_to_gguf.py @@ -13,7 +13,7 @@ import sys from enum import IntEnum from pathlib import Path from hashlib import sha256 -from typing import TYPE_CHECKING, Any, Callable, ContextManager, Iterable, Iterator, Sequence, TypeVar, cast +from typing import TYPE_CHECKING, Any, Callable, ContextManager, Iterable, Iterator, Literal, Sequence, TypeVar, cast import math import numpy as np @@ -404,7 +404,7 @@ class Model: return tokens, toktypes, tokpre - # NOTE: this function is generated by convert-hf-to-gguf-update.py + # NOTE: this function is generated by convert_hf_to_gguf_update.py # do not modify it manually! # ref: https://github.com/ggerganov/llama.cpp/pull/6920 # Marker: Start get_vocab_base_pre @@ -424,7 +424,7 @@ class Model: res = None - # NOTE: if you get an error here, you need to update the convert-hf-to-gguf-update.py script + # NOTE: if you get an error here, you need to update the convert_hf_to_gguf_update.py script # or pull the latest version of the model from Huggingface # don't edit the hashes manually! if chkhsh == "0ef9807a4087ebef797fc749390439009c3b9eda9ad1a097abbe738f486c01e5": @@ -502,9 +502,9 @@ class Model: logger.warning("**************************************************************************************") logger.warning("** WARNING: The BPE pre-tokenizer was not recognized!") logger.warning("** There are 2 possible reasons for this:") - logger.warning("** - the model has not been added to convert-hf-to-gguf-update.py yet") + logger.warning("** - the model has not been added to convert_hf_to_gguf_update.py yet") logger.warning("** - the pre-tokenization config has changed upstream") - logger.warning("** Check your model files and convert-hf-to-gguf-update.py and update them accordingly.") + logger.warning("** Check your model files and convert_hf_to_gguf_update.py and update them accordingly.") logger.warning("** ref: https://github.com/ggerganov/llama.cpp/pull/6920") logger.warning("**") logger.warning(f"** chkhsh: {chkhsh}") @@ -680,6 +680,51 @@ class Model: special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens)) special_vocab.add_to_gguf(self.gguf_writer) + def _set_vocab_builtin(self, model_name: Literal["gpt-neox", "llama-spm"], vocab_size: int): + tokenizer_path = Path(sys.path[0]) / "models" / f"ggml-vocab-{model_name}.gguf" + logger.warning(f"Using tokenizer from '{os.path.relpath(tokenizer_path, os.getcwd())}'") + vocab_reader = gguf.GGUFReader(tokenizer_path, "r") + + default_pre = "mpt" if model_name == "gpt-neox" else "default" + + field = vocab_reader.get_field(gguf.Keys.Tokenizer.MODEL) + assert field # tokenizer model + self.gguf_writer.add_tokenizer_model(bytes(field.parts[-1]).decode("utf-8")) + + field = vocab_reader.get_field(gguf.Keys.Tokenizer.PRE) + self.gguf_writer.add_tokenizer_pre(bytes(field.parts[-1]).decode("utf-8") if field else default_pre) + + field = vocab_reader.get_field(gguf.Keys.Tokenizer.LIST) + assert field # token list + self.gguf_writer.add_token_list([bytes(field.parts[i]) for i in field.data][:vocab_size]) + + if model_name == "llama-spm": + field = vocab_reader.get_field(gguf.Keys.Tokenizer.SCORES) + assert field # token scores + self.gguf_writer.add_token_scores([field.parts[i].tolist()[0] for i in field.data][:vocab_size]) + + field = vocab_reader.get_field(gguf.Keys.Tokenizer.TOKEN_TYPE) + assert field # token types + self.gguf_writer.add_token_types([field.parts[i].tolist()[0] for i in field.data][:vocab_size]) + + if model_name != "llama-spm": + field = vocab_reader.get_field(gguf.Keys.Tokenizer.MERGES) + assert field # token merges + self.gguf_writer.add_token_merges([bytes(field.parts[i]) for i in field.data]) + + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.BOS_ID)) is not None: + self.gguf_writer.add_bos_token_id(field.parts[-1].tolist()[0]) + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.EOS_ID)) is not None: + self.gguf_writer.add_eos_token_id(field.parts[-1].tolist()[0]) + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.UNK_ID)) is not None: + self.gguf_writer.add_unk_token_id(field.parts[-1].tolist()[0]) + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.PAD_ID)) is not None: + self.gguf_writer.add_pad_token_id(field.parts[-1].tolist()[0]) + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.ADD_BOS)) is not None: + self.gguf_writer.add_add_bos_token(field.parts[-1].tolist()[0]) + if (field := vocab_reader.get_field(gguf.Keys.Tokenizer.ADD_EOS)) is not None: + self.gguf_writer.add_add_eos_token(field.parts[-1].tolist()[0]) + @Model.register("GPTNeoXForCausalLM") class GPTNeoXModel(Model): @@ -1945,7 +1990,7 @@ class Phi3MiniModel(Model): if len(rope_scaling_type) == 0: raise KeyError('Missing the required key rope_scaling.type') - if rope_scaling_type == 'su': + if rope_scaling_type == 'su' or rope_scaling_type == 'longrope': attn_factor = math.sqrt(1 + math.log(scale) / math.log(orig_max_pos_embds)) if scale > 1.0 else 1.0 elif rope_scaling_type == 'yarn': attn_factor = 0.1 * math.log(scale) + 1.0 if scale > 1.0 else 1.0 @@ -2319,6 +2364,8 @@ class GemmaModel(Model): special_vocab._set_special_token("eot", 107) special_vocab.add_to_gguf(self.gguf_writer) + self.gguf_writer.add_add_space_prefix(False) + def set_gguf_parameters(self): hparams = self.hparams block_count = hparams["num_hidden_layers"] @@ -2369,6 +2416,7 @@ class Gemma2Model(Model): special_vocab = gguf.SpecialVocab(self.dir_model, n_vocab=len(tokens)) special_vocab.add_to_gguf(self.gguf_writer) + self.gguf_writer.add_add_space_prefix(False) def set_gguf_parameters(self): @@ -2400,7 +2448,7 @@ class Gemma2Model(Model): raise ValueError("query_pre_attn_scalar must be equal to n_embd / n_head") def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: - del bid # unusem + del bid # unused # lm_head is not used in llama.cpp, while autoawq will include this tensor in model # To prevent errors, skip loading lm_head.weight. @@ -2439,39 +2487,7 @@ class MambaModel(Model): self._set_vocab_sentencepiece() else: # Use the GPT-NeoX tokenizer when no tokenizer files are present - tokenizer_path = Path(sys.path[0]) / "models" / "ggml-vocab-gpt-neox.gguf" - logger.warning(f"Using tokenizer from '{os.path.relpath(tokenizer_path, os.getcwd())}'") - neox_reader = gguf.GGUFReader(tokenizer_path, "r") - - field = neox_reader.get_field(gguf.Keys.Tokenizer.MODEL) - self.gguf_writer.add_tokenizer_model(bytes(field.parts[-1]).decode("utf-8") if field else "gpt2") - - field = neox_reader.get_field(gguf.Keys.Tokenizer.PRE) - self.gguf_writer.add_tokenizer_pre(bytes(field.parts[-1]).decode("utf-8") if field else "mpt") - - field = neox_reader.get_field(gguf.Keys.Tokenizer.LIST) - assert field - self.gguf_writer.add_token_list([bytes(field.parts[i]) for i in field.data][:vocab_size]) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.TOKEN_TYPE) - assert field - self.gguf_writer.add_token_types([field.parts[i].tolist()[0] for i in field.data][:vocab_size]) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.MERGES) - assert field - self.gguf_writer.add_token_merges([bytes(field.parts[i]) for i in field.data]) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.BOS_ID) - self.gguf_writer.add_bos_token_id(field.parts[-1].tolist()[0] if field else 1) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.EOS_ID) - self.gguf_writer.add_eos_token_id(field.parts[-1].tolist()[0] if field else 0) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.UNK_ID) - self.gguf_writer.add_unk_token_id(field.parts[-1].tolist()[0] if field else 0) - - field = neox_reader.get_field(gguf.Keys.Tokenizer.PAD_ID) - self.gguf_writer.add_pad_token_id(field.parts[-1].tolist()[0] if field else 0) + self._set_vocab_builtin("gpt-neox", vocab_size) def set_gguf_parameters(self): d_model = self.find_hparam(["hidden_size", "d_model"]) @@ -2623,6 +2639,82 @@ class JinaBertV2Model(BertModel): self.gguf_writer.add_add_eos_token(True) +@Model.register("OpenELMForCausalLM") +class OpenELMModel(Model): + model_arch = gguf.MODEL_ARCH.OPENELM + + @staticmethod + def _make_divisible(v: float | int, divisor: int) -> int: + # ref: https://huggingface.co/apple/OpenELM-270M-Instruct/blob/eb111ff2e6724348e5b905984063d4064d4bc579/configuration_openelm.py#L34-L38 + new_v = max(divisor, int(v + divisor / 2) // divisor * divisor) + # Make sure that round down does not go down by more than 10%. + if new_v < 0.9 * v: + new_v += divisor + return new_v + + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + + ffn_multipliers: list[float] = self.hparams["ffn_multipliers"] + ffn_dim_divisor: int = self.hparams["ffn_dim_divisor"] + self._n_embd: int = self.hparams["model_dim"] + self._num_kv_heads: list[int] = self.hparams["num_kv_heads"] + self._num_query_heads: list[int] = self.hparams["num_query_heads"] + self._ffn_dims: list[int] = [ + OpenELMModel._make_divisible(multiplier * self._n_embd, ffn_dim_divisor) + for multiplier in ffn_multipliers + ] + assert isinstance(self._num_kv_heads, list) and isinstance(self._num_kv_heads[0], int) + assert isinstance(self._num_query_heads, list) and isinstance(self._num_query_heads[0], int) + + # Uses the tokenizer from meta-llama/Llama-2-7b-hf + def set_vocab(self): + try: + self._set_vocab_sentencepiece() + except FileNotFoundError: + self._set_vocab_builtin("llama-spm", self.hparams["vocab_size"]) + + def set_gguf_parameters(self): + n_embd = self._n_embd + head_dim = self.hparams["head_dim"] + rot_pct = 1.0 + assert self.block_count == len(self._num_kv_heads) + assert self.block_count == len(self._num_query_heads) + assert self.block_count == len(self._ffn_dims) + + self.gguf_writer.add_name(self.dir_model.name if self.model_name is None else self.model_name) + self.gguf_writer.add_block_count(self.block_count) + self.gguf_writer.add_context_length(self.hparams["max_context_length"]) + self.gguf_writer.add_embedding_length(n_embd) + self.gguf_writer.add_feed_forward_length(self._ffn_dims) + self.gguf_writer.add_head_count(self._num_query_heads) + self.gguf_writer.add_head_count_kv(self._num_kv_heads) + self.gguf_writer.add_rope_freq_base(self.hparams["rope_freq_constant"]) + # https://huggingface.co/apple/OpenELM-270M-Instruct/blob/c401df2/modeling_openelm.py#L30 + self.gguf_writer.add_layer_norm_rms_eps(1e-6) + self.gguf_writer.add_rope_dimension_count(int(rot_pct * head_dim)) + self.gguf_writer.add_key_length(head_dim) + self.gguf_writer.add_value_length(head_dim) + self.gguf_writer.add_file_type(self.ftype) + + def find_hparam(self, keys: Iterable[str], optional: bool = False) -> Any: + if "n_layers" in keys: + return self.hparams["num_transformer_layers"] + + return super().find_hparam(keys, optional) + + def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: + + # split ff + if bid is not None and name == f"transformer.layers.{bid}.ffn.proj_1.weight": + ff_dim = self._ffn_dims[bid] + yield (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_GATE, bid), data_torch[:ff_dim]) + yield (self.format_tensor_name(gguf.MODEL_TENSOR.FFN_UP, bid), data_torch[ff_dim:]) + return + + yield (self.map_tensor_name(name), data_torch) + + @Model.register("ArcticForCausalLM") class ArcticModel(Model): model_arch = gguf.MODEL_ARCH.ARCTIC @@ -2853,11 +2945,17 @@ class DeepseekV2Model(Model): raise ValueError(f"Unprocessed experts: {experts}") -@Model.register("T5ForConditionalGeneration") @Model.register("T5WithLMHeadModel") +@Model.register("T5ForConditionalGeneration") +@Model.register("MT5ForConditionalGeneration") +@Model.register("UMT5ForConditionalGeneration") class T5Model(Model): model_arch = gguf.MODEL_ARCH.T5 + def __init__(self, *args, **kwargs): + super().__init__(*args, **kwargs) + self.shared_token_embeddings_found = False + def set_vocab(self): # to avoid TypeError: Descriptors cannot be created directly # exception when importing sentencepiece_model_pb2 @@ -2865,17 +2963,29 @@ class T5Model(Model): from sentencepiece import SentencePieceProcessor from sentencepiece import sentencepiece_model_pb2 as model - tokenizer_path = self.dir_model / 'spiece.model' + tokenizer_path = self.dir_model / 'tokenizer.model' + + # many older models use spiece.model tokenizer model filename + if not tokenizer_path.is_file(): + tokenizer_path = self.dir_model / 'spiece.model' if not tokenizer_path.is_file(): raise FileNotFoundError(f"File not found: {tokenizer_path}") sentencepiece_model = model.ModelProto() sentencepiece_model.ParseFromString(open(tokenizer_path, "rb").read()) + + # some models like Pile-T5 family use BPE tokenizer instead of Unigram + if sentencepiece_model.trainer_spec.model_type == 2: # BPE + # assure the tokenizer model file name is correct + assert tokenizer_path.name == 'tokenizer.model' + return self._set_vocab_sentencepiece() + else: + assert sentencepiece_model.trainer_spec.model_type == 1 # UNIGRAM + add_prefix = sentencepiece_model.normalizer_spec.add_dummy_prefix remove_whitespaces = sentencepiece_model.normalizer_spec.remove_extra_whitespaces precompiled_charsmap = sentencepiece_model.normalizer_spec.precompiled_charsmap - assert sentencepiece_model.trainer_spec.model_type == 1 # UNIGRAM tokenizer = SentencePieceProcessor() tokenizer.LoadFromFile(str(tokenizer_path)) @@ -2945,7 +3055,10 @@ class T5Model(Model): def set_gguf_parameters(self): self.gguf_writer.add_name("T5") - self.gguf_writer.add_context_length(self.hparams["n_positions"]) + if (n_ctx := self.find_hparam(["n_positions"], optional=True)) is None: + logger.warning("Couldn't find context length in config.json, assuming default value of 512") + n_ctx = 512 + self.gguf_writer.add_context_length(n_ctx) self.gguf_writer.add_embedding_length(self.hparams["d_model"]) self.gguf_writer.add_feed_forward_length(self.hparams["d_ff"]) self.gguf_writer.add_block_count(self.hparams["num_layers"]) @@ -2961,12 +3074,17 @@ class T5Model(Model): def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]: del bid # unused - # Sometimes T5 and Flan-T5 based models contain "encoder.embed_tokens.weight" tensor or - # "decoder.embed_tokens.weight" tensors that are duplicates of "shared.weight" tensor - # To prevent errors caused by an unnecessary unmapped tensor, skip both of them and use only "shared.weight". - if name == "decoder.embed_tokens.weight" or name == "encoder.embed_tokens.weight": - logger.debug(f"Skipping tensor {name!r} in safetensors so that convert can end normally.") - return [] + # T5 based models contain shared token embeddings tensors saved randomly as either "encoder.embed_tokens.weight", + # "decoder.embed_tokens.weight" or "shared.weight" tensor. In some models there are even multiple of them stored + # in the safetensors files. We use the first tensor from these three as the token embeddings for both encoder + # and decoder and ignore the remaining ones. + if name in ["decoder.embed_tokens.weight", "encoder.embed_tokens.weight", "shared.weight"]: + if not self.shared_token_embeddings_found: + name = "shared.weight" + self.shared_token_embeddings_found = True + else: + logger.debug(f"Skipping shared tensor {name!r} in safetensors so that convert can end normally.") + return [] return [(self.map_tensor_name(name), data_torch)] @@ -3294,10 +3412,6 @@ def parse_args() -> argparse.Namespace: "--vocab-only", action="store_true", help="extract only the vocab", ) - parser.add_argument( - "--awq-path", type=Path, default=None, - help="Path to scale awq cache file", - ) parser.add_argument( "--outfile", type=Path, help="path to write to; default: based on input. {ftype} will be replaced by the outtype.", @@ -3375,19 +3489,6 @@ def main() -> None: dir_model = args.model - if args.awq_path: - sys.path.insert(1, str(Path(__file__).parent / 'awq-py')) - from awq.apply_awq import add_scale_weights # type: ignore[import-not-found] - tmp_model_path = args.model / "weighted_model" - dir_model = tmp_model_path - if tmp_model_path.is_dir(): - logger.info(f"{tmp_model_path} exists as a weighted model.") - else: - tmp_model_path.mkdir(parents=True, exist_ok=True) - logger.info("Saving new weighted model ...") - add_scale_weights(str(args.model), str(args.awq_path), str(tmp_model_path)) - logger.info(f"Saved weighted model at {tmp_model_path}.") - if not dir_model.is_dir(): logger.error(f'Error: {args.model} is not a directory') sys.exit(1) diff --git a/convert-hf-to-gguf-update.py b/convert_hf_to_gguf_update.py similarity index 87% rename from convert-hf-to-gguf-update.py rename to convert_hf_to_gguf_update.py index 944e9d15a..e4165ae2d 100755 --- a/convert-hf-to-gguf-update.py +++ b/convert_hf_to_gguf_update.py @@ -2,7 +2,7 @@ # -*- coding: utf-8 -*- # This script downloads the tokenizer models of the specified models from Huggingface and -# generates the get_vocab_base_pre() function for convert-hf-to-gguf.py +# generates the get_vocab_base_pre() function for convert_hf_to_gguf.py # # This is necessary in order to analyze the type of pre-tokenizer used by the model and # provide the necessary information to llama.cpp via the GGUF header in order to implement @@ -15,9 +15,9 @@ # - Add a new model to the "models" list # - Run the script with your huggingface token: # -# python3 convert-hf-to-gguf-update.py +# python3 convert_hf_to_gguf_update.py # -# - Copy-paste the generated get_vocab_base_pre() function into convert-hf-to-gguf.py +# - Copy-paste the generated get_vocab_base_pre() function into convert_hf_to_gguf.py # - Update llama.cpp with the new pre-tokenizer if necessary # # TODO: generate tokenizer tests for llama.cpp @@ -37,7 +37,7 @@ from enum import IntEnum, auto from transformers import AutoTokenizer logging.basicConfig(level=logging.DEBUG) -logger = logging.getLogger("convert-hf-to-gguf-update") +logger = logging.getLogger("convert_hf_to_gguf_update") sess = requests.Session() @@ -45,6 +45,7 @@ class TOKENIZER_TYPE(IntEnum): SPM = auto() BPE = auto() WPM = auto() + UGM = auto() # TODO: this string has to exercise as much pre-tokenizer functionality as possible @@ -55,10 +56,10 @@ if len(sys.argv) == 2: token = sys.argv[1] if not token.startswith("hf_"): logger.info("Huggingface token seems invalid") - logger.info("Usage: python convert-hf-to-gguf-update.py ") + logger.info("Usage: python convert_hf_to_gguf_update.py ") sys.exit(1) else: - logger.info("Usage: python convert-hf-to-gguf-update.py ") + logger.info("Usage: python convert_hf_to_gguf_update.py ") sys.exit(1) # TODO: add models here, base models preferred @@ -86,7 +87,10 @@ models = [ {"name": "poro-chat", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Poro-34B-chat", }, {"name": "jina-v2-code", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/jinaai/jina-embeddings-v2-base-code", }, {"name": "viking", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LumiOpen/Viking-7B", }, # Also used for Viking 13B and 33B + {"name": "gemma", "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/google/gemma-2b", }, + {"name": "gemma-2", "tokt": TOKENIZER_TYPE.SPM, "repo": "https://huggingface.co/google/gemma-2-9b", }, {"name": "jais", "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/core42/jais-13b", }, + {"name": "t5", "tokt": TOKENIZER_TYPE.UGM, "repo": "https://huggingface.co/google-t5/t5-small", }, ] @@ -108,9 +112,13 @@ def download_model(model): os.makedirs(f"models/tokenizers/{name}", exist_ok=True) files = ["config.json", "tokenizer.json", "tokenizer_config.json"] + if tokt == TOKENIZER_TYPE.SPM: files.append("tokenizer.model") + if tokt == TOKENIZER_TYPE.UGM: + files.append("spiece.model") + for file in files: save_path = f"models/tokenizers/{name}/{file}" if os.path.isfile(save_path): @@ -126,14 +134,14 @@ for model in models: logger.error(f"Failed to download model {model['name']}. Error: {e}") -# generate the source code for the convert-hf-to-gguf.py:get_vocab_base_pre() function: +# generate the source code for the convert_hf_to_gguf.py:get_vocab_base_pre() function: src_ifs = "" for model in models: name = model["name"] tokt = model["tokt"] - if tokt == TOKENIZER_TYPE.SPM: + if tokt == TOKENIZER_TYPE.SPM or tokt == TOKENIZER_TYPE.UGM: continue # Skip if the tokenizer folder does not exist or there are other download issues previously @@ -143,7 +151,10 @@ for model in models: # create the tokenizer try: - tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}") + if name == "t5": + tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}", use_fast=False) + else: + tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}") except OSError as e: logger.error(f"Error loading tokenizer for model {name}. The model may not exist or is not accessible with the provided token. Error: {e}") continue # Skip to the next model if the tokenizer can't be loaded @@ -190,7 +201,7 @@ src_func = f""" res = None - # NOTE: if you get an error here, you need to update the convert-hf-to-gguf-update.py script + # NOTE: if you get an error here, you need to update the convert_hf_to_gguf_update.py script # or pull the latest version of the model from Huggingface # don't edit the hashes manually! {src_ifs} @@ -199,9 +210,9 @@ src_func = f""" logger.warning("**************************************************************************************") logger.warning("** WARNING: The BPE pre-tokenizer was not recognized!") logger.warning("** There are 2 possible reasons for this:") - logger.warning("** - the model has not been added to convert-hf-to-gguf-update.py yet") + logger.warning("** - the model has not been added to convert_hf_to_gguf_update.py yet") logger.warning("** - the pre-tokenization config has changed upstream") - logger.warning("** Check your model files and convert-hf-to-gguf-update.py and update them accordingly.") + logger.warning("** Check your model files and convert_hf_to_gguf_update.py and update them accordingly.") logger.warning("** ref: https://github.com/ggerganov/llama.cpp/pull/6920") logger.warning("**") logger.warning(f"** chkhsh: {{chkhsh}}") @@ -215,7 +226,7 @@ src_func = f""" return res """ -convert_py_pth = pathlib.Path("convert-hf-to-gguf.py") +convert_py_pth = pathlib.Path("convert_hf_to_gguf.py") convert_py = convert_py_pth.read_text(encoding="utf-8") convert_py = re.sub( r"(# Marker: Start get_vocab_base_pre)(.+?)( +# Marker: End get_vocab_base_pre)", @@ -226,7 +237,7 @@ convert_py = re.sub( convert_py_pth.write_text(convert_py, encoding="utf-8") -logger.info("+++ convert-hf-to-gguf.py was updated") +logger.info("+++ convert_hf_to_gguf.py was updated") # generate tests for each tokenizer model @@ -264,6 +275,7 @@ tests = [ "\n =", "' era", "Hello, y'all! How are you 😁 ?我想在apple工作1314151天～", + "!!!!!!", "3", "33", "333", @@ -273,7 +285,8 @@ tests = [ "3333333", "33333333", "333333333", - # "Cửa Việt", # llama-bpe fails on this + "Cửa Việt", # llama-bpe fails on this + " discards", chktxt, ] @@ -301,7 +314,10 @@ for model in models: # create the tokenizer try: - tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}") + if name == "t5": + tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}", use_fast=False) + else: + tokenizer = AutoTokenizer.from_pretrained(f"models/tokenizers/{name}") except OSError as e: logger.error(f"Failed to load tokenizer for model {name}. Error: {e}") continue # Skip this model and continue with the next one in the loop @@ -327,6 +343,6 @@ logger.info("\nRun the following commands to generate the vocab files for testin for model in models: name = model["name"] - print(f"python3 convert-hf-to-gguf.py models/tokenizers/{name}/ --outfile models/ggml-vocab-{name}.gguf --vocab-only") # noqa: NP100 + print(f"python3 convert_hf_to_gguf.py models/tokenizers/{name}/ --outfile models/ggml-vocab-{name}.gguf --vocab-only") # noqa: NP100 logger.info("\n") diff --git a/convert-llama-ggml-to-gguf.py b/convert_llama_ggml_to_gguf.py similarity index 100% rename from convert-llama-ggml-to-gguf.py rename to convert_llama_ggml_to_gguf.py diff --git a/docs/android.md b/docs/android.md new file mode 100644 index 000000000..cec4358d9 --- /dev/null +++ b/docs/android.md @@ -0,0 +1,56 @@ + +# Android + +## Build on Android using Termux +[Termux](https://github.com/termux/termux-app#installation) is a method to execute `llama.cpp` on an Android device (no root required). +``` +apt update && apt upgrade -y +apt install git make cmake +``` + +It's recommended to move your model inside the `~/` directory for best performance: +``` +cd storage/downloads +mv model.gguf ~/ +``` + +[Get the code](https://github.com/ggerganov/llama.cpp#get-the-code) & [follow the Linux build instructions](https://github.com/ggerganov/llama.cpp#build) to build `llama.cpp`. + +## Building the Project using Android NDK +Obtain the [Android NDK](https://developer.android.com/ndk) and then build with CMake. + +Execute the following commands on your computer to avoid downloading the NDK to your mobile. Alternatively, you can also do this in Termux: +``` +$ mkdir build-android +$ cd build-android +$ export NDK= +$ cmake -DCMAKE_TOOLCHAIN_FILE=$NDK/build/cmake/android.toolchain.cmake -DANDROID_ABI=arm64-v8a -DANDROID_PLATFORM=android-23 -DCMAKE_C_FLAGS=-march=armv8.4a+dotprod .. +$ make +``` + +Install [termux](https://github.com/termux/termux-app#installation) on your device and run `termux-setup-storage` to get access to your SD card (if Android 11+ then run the command twice). + +Finally, copy these built `llama` binaries and the model file to your device storage. Because the file permissions in the Android sdcard cannot be changed, you can copy the executable files to the `/data/data/com.termux/files/home/bin` path, and then execute the following commands in Termux to add executable permission: + +(Assumed that you have pushed the built executable files to the /sdcard/llama.cpp/bin path using `adb push`) +``` +$cp -r /sdcard/llama.cpp/bin /data/data/com.termux/files/home/ +$cd /data/data/com.termux/files/home/bin +$chmod +x ./* +``` + +Download model [llama-2-7b-chat.Q4_K_M.gguf](https://huggingface.co/TheBloke/Llama-2-7B-Chat-GGUF/blob/main/llama-2-7b-chat.Q4_K_M.gguf), and push it to `/sdcard/llama.cpp/`, then move it to `/data/data/com.termux/files/home/model/` + +``` +$mv /sdcard/llama.cpp/llama-2-7b-chat.Q4_K_M.gguf /data/data/com.termux/files/home/model/ +``` + +Now, you can start chatting: +``` +$cd /data/data/com.termux/files/home/bin +$./llama-cli -m ../model/llama-2-7b-chat.Q4_K_M.gguf -n 128 -cml +``` + +Here's a demo of an interactive session running on Pixel 5 phone: + +https://user-images.githubusercontent.com/271616/225014776-1d567049-ad71-4ef2-b050-55b0b3b9274c.mp4 diff --git a/docs/BLIS.md b/docs/backend/BLIS.md similarity index 100% rename from docs/BLIS.md rename to docs/backend/BLIS.md diff --git a/README-sycl.md b/docs/backend/SYCL.md similarity index 100% rename from README-sycl.md rename to docs/backend/SYCL.md diff --git a/docs/build.md b/docs/build.md new file mode 100644 index 000000000..bf41bfdf9 --- /dev/null +++ b/docs/build.md @@ -0,0 +1,288 @@ +# Build llama.cpp locally + +**To get the Code:** + +```bash +git clone https://github.com/ggerganov/llama.cpp +cd llama.cpp +``` + +In order to build llama.cpp you have four different options. + +- Using `make`: + - On Linux or MacOS: + + ```bash + make + ``` + + - On Windows: + + 1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases). + 2. Extract `w64devkit` on your pc. + 3. Run `w64devkit.exe`. + 4. Use the `cd` command to reach the `llama.cpp` folder. + 5. From here you can run: + ```bash + make + ``` + + - Notes: + - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `make -j 8` will run 8 jobs in parallel. + - For faster repeated compilation, install [ccache](https://ccache.dev/). + - For debug builds, run `make LLAMA_DEBUG=1` + +- Using `CMake`: + + ```bash + cmake -B build + cmake --build build --config Release + ``` + + **Notes**: + + - For faster compilation, add the `-j` argument to run multiple jobs in parallel. For example, `cmake --build build --config Release -j 8` will run 8 jobs in parallel. + - For faster repeated compilation, install [ccache](https://ccache.dev/). + - For debug builds, there are two cases: + + 1. Single-config generators (e.g. default = `Unix Makefiles`; note that they just ignore the `--config` flag): + + ```bash + cmake -B build -DCMAKE_BUILD_TYPE=Debug + cmake --build build + ``` + + 2. Multi-config generators (`-G` param set to Visual Studio, XCode...): + + ```bash + cmake -B build -G "Xcode" + cmake --build build --config Debug + ``` + +- Using `gmake` (FreeBSD): + + 1. Install and activate [DRM in FreeBSD](https://wiki.freebsd.org/Graphics) + 2. Add your user to **video** group + 3. Install compilation dependencies. + + ```bash + sudo pkg install gmake automake autoconf pkgconf llvm15 openblas + + gmake CC=/usr/local/bin/clang15 CXX=/usr/local/bin/clang++15 -j4 + ``` + +## Metal Build + +On MacOS, Metal is enabled by default. Using Metal makes the computation run on the GPU. +To disable the Metal build at compile time use the `GGML_NO_METAL=1` flag or the `GGML_METAL=OFF` cmake option. + +When built with Metal support, you can explicitly disable GPU inference with the `--n-gpu-layers|-ngl 0` command-line +argument. + +## BLAS Build + +Building the program with BLAS support may lead to some performance improvements in prompt processing using batch sizes higher than 32 (the default is 512). Support with CPU-only BLAS implementations doesn't affect the normal generation performance. We may see generation performance improvements with GPU-involved BLAS implementations, e.g. cuBLAS, hipBLAS. There are currently several different BLAS implementations available for build and use: + +### Accelerate Framework: + +This is only available on Mac PCs and it's enabled by default. You can just build using the normal instructions. + +### OpenBLAS: + +This provides BLAS acceleration using only the CPU. Make sure to have OpenBLAS installed on your machine. + +- Using `make`: + - On Linux: + ```bash + make GGML_OPENBLAS=1 + ``` + + - On Windows: + + 1. Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases). + 2. Download the latest version of [OpenBLAS for Windows](https://github.com/xianyi/OpenBLAS/releases). + 3. Extract `w64devkit` on your pc. + 4. From the OpenBLAS zip that you just downloaded copy `libopenblas.a`, located inside the `lib` folder, inside `w64devkit\x86_64-w64-mingw32\lib`. + 5. From the same OpenBLAS zip copy the content of the `include` folder inside `w64devkit\x86_64-w64-mingw32\include`. + 6. Run `w64devkit.exe`. + 7. Use the `cd` command to reach the `llama.cpp` folder. + 8. From here you can run: + + ```bash + make GGML_OPENBLAS=1 + ``` + +- Using `CMake` on Linux: + + ```bash + cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=OpenBLAS + cmake --build build --config Release + ``` + +### BLIS + +Check [BLIS.md](./backend/BLIS.md) for more information. + +### SYCL + +SYCL is a higher-level programming model to improve programming productivity on various hardware accelerators. + +llama.cpp based on SYCL is used to **support Intel GPU** (Data Center Max series, Flex series, Arc series, Built-in GPU and iGPU). + +For detailed info, please refer to [llama.cpp for SYCL](./backend/SYCL.md). + +### Intel oneMKL + +Building through oneAPI compilers will make avx_vnni instruction set available for intel processors that do not support avx512 and avx512_vnni. Please note that this build config **does not support Intel GPU**. For Intel GPU support, please refer to [llama.cpp for SYCL](./backend/SYCL.md). + +- Using manual oneAPI installation: + By default, `GGML_BLAS_VENDOR` is set to `Generic`, so if you already sourced intel environment script and assign `-DGGML_BLAS=ON` in cmake, the mkl version of Blas will automatically been selected. Otherwise please install oneAPI and follow the below steps: + ```bash + source /opt/intel/oneapi/setvars.sh # You can skip this step if in oneapi-basekit docker image, only required for manual installation + cmake -B build -DGGML_BLAS=ON -DGGML_BLAS_VENDOR=Intel10_64lp -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DGGML_NATIVE=ON + cmake --build build --config Release + ``` + +- Using oneAPI docker image: + If you do not want to source the environment vars and install oneAPI manually, you can also build the code using intel docker container: [oneAPI-basekit](https://hub.docker.com/r/intel/oneapi-basekit). Then, you can use the commands given above. + +Check [Optimizing and Running LLaMA2 on Intel® CPU](https://www.intel.com/content/www/us/en/content-details/791610/optimizing-and-running-llama2-on-intel-cpu.html) for more information. + +### CUDA + +This provides GPU acceleration using the CUDA cores of your Nvidia GPU. Make sure to have the CUDA toolkit installed. You can download it from your Linux distro's package manager (e.g. `apt install nvidia-cuda-toolkit`) or from here: [CUDA Toolkit](https://developer.nvidia.com/cuda-downloads). + +For Jetson user, if you have Jetson Orin, you can try this: [Offical Support](https://www.jetson-ai-lab.com/tutorial_text-generation.html). If you are using an old model(nano/TX2), need some additional operations before compiling. + +- Using `make`: + ```bash + make GGML_CUDA=1 + ``` +- Using `CMake`: + + ```bash + cmake -B build -DGGML_CUDA=ON + cmake --build build --config Release + ``` + +The environment variable [`CUDA_VISIBLE_DEVICES`](https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html#env-vars) can be used to specify which GPU(s) will be used. The following compilation options are also available to tweak performance: + +| Option | Legal values | Default | Description | +|-------------------------------|------------------------|---------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| GGML_CUDA_FORCE_DMMV | Boolean | false | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 6.1/Pascal/GTX 1000 or higher). Does not affect k-quants. | +| GGML_CUDA_DMMV_X | Positive integer >= 32 | 32 | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. | +| GGML_CUDA_MMV_Y | Positive integer | 1 | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. | +| GGML_CUDA_FORCE_MMQ | Boolean | false | Force the use of custom matrix multiplication kernels for quantized models instead of FP16 cuBLAS even if there is no int8 tensor core implementation available (affects V100, RDNA3). MMQ kernels are enabled by default on GPUs with int8 tensor core support. With MMQ force enabled, speed for large batch sizes will be worse but VRAM consumption will be lower. | +| GGML_CUDA_FORCE_CUBLAS | Boolean | false | Force the use of FP16 cuBLAS instead of custom matrix multiplication kernels for quantized models | +| GGML_CUDA_F16 | Boolean | false | If enabled, use half-precision floating point arithmetic for the CUDA dequantization + mul mat vec kernels and for the q4_1 and q5_1 matrix matrix multiplication kernels. Can improve performance on relatively recent GPUs. | +| GGML_CUDA_KQUANTS_ITER | 1 or 2 | 2 | Number of values processed per iteration and per CUDA thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs. | +| GGML_CUDA_PEER_MAX_BATCH_SIZE | Positive integer | 128 | Maximum batch size for which to enable peer access between multiple GPUs. Peer access requires either Linux or NVLink. When using NVLink enabling peer access for larger batch sizes is potentially beneficial. | +| GGML_CUDA_FA_ALL_QUANTS | Boolean | false | Compile support for all KV cache quantization type (combinations) for the FlashAttention CUDA kernels. More fine-grained control over KV cache size but compilation takes much longer. | + +### hipBLAS + +This provides BLAS acceleration on HIP-supported AMD GPUs. +Make sure to have ROCm installed. +You can download it from your Linux distro's package manager or from here: [ROCm Quick Start (Linux)](https://rocm.docs.amd.com/projects/install-on-linux/en/latest/tutorial/quick-start.html#rocm-install-quick). + +- Using `make`: + ```bash + make GGML_HIPBLAS=1 + ``` +- Using `CMake` for Linux (assuming a gfx1030-compatible AMD GPU): + ```bash + HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -R)" \ + cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \ + && cmake --build build --config Release -- -j 16 + ``` + On Linux it is also possible to use unified memory architecture (UMA) to share main memory between the CPU and integrated GPU by setting `-DGGML_HIP_UMA=ON`. + However, this hurts performance for non-integrated GPUs (but enables working with integrated GPUs). + + Note that if you get the following error: + ``` + clang: error: cannot find ROCm device library; provide its path via '--rocm-path' or '--rocm-device-lib-path', or pass '-nogpulib' to build without ROCm device library + ``` + Try searching for a directory under `HIP_PATH` that contains the file + `oclc_abi_version_400.bc`. Then, add the following to the start of the + command: `HIP_DEVICE_LIB_PATH=`, so something + like: + ```bash + HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \ + HIP_DEVICE_LIB_PATH= \ + cmake -S . -B build -DGGML_HIPBLAS=ON -DAMDGPU_TARGETS=gfx1030 -DCMAKE_BUILD_TYPE=Release \ + && cmake --build build -- -j 16 + ``` + +- Using `make` (example for target gfx1030, build with 16 CPU threads): + ```bash + make -j16 GGML_HIPBLAS=1 GGML_HIP_UMA=1 AMDGPU_TARGETS=gfx1030 + ``` + +- Using `CMake` for Windows (using x64 Native Tools Command Prompt for VS, and assuming a gfx1100-compatible AMD GPU): + ```bash + set PATH=%HIP_PATH%\bin;%PATH% + cmake -S . -B build -G Ninja -DAMDGPU_TARGETS=gfx1100 -DGGML_HIPBLAS=ON -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ -DCMAKE_BUILD_TYPE=Release + cmake --build build + ``` + Make sure that `AMDGPU_TARGETS` is set to the GPU arch you want to compile for. The above example uses `gfx1100` that corresponds to Radeon RX 7900XTX/XT/GRE. You can find a list of targets [here](https://llvm.org/docs/AMDGPUUsage.html#processors) + Find your gpu version string by matching the most significant version information from `rocminfo | grep gfx | head -1 | awk '{print $2}'` with the list of processors, e.g. `gfx1035` maps to `gfx1030`. + + +The environment variable [`HIP_VISIBLE_DEVICES`](https://rocm.docs.amd.com/en/latest/understand/gpu_isolation.html#hip-visible-devices) can be used to specify which GPU(s) will be used. +If your GPU is not officially supported you can use the environment variable [`HSA_OVERRIDE_GFX_VERSION`] set to a similar GPU, for example 10.3.0 on RDNA2 (e.g. gfx1030, gfx1031, or gfx1035) or 11.0.0 on RDNA3. +The following compilation options are also available to tweak performance (yes, they refer to CUDA, not HIP, because it uses the same code as the cuBLAS version above): + +| Option | Legal values | Default | Description | +|------------------------|------------------------|---------|------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------| +| GGML_CUDA_DMMV_X | Positive integer >= 32 | 32 | Number of values in x direction processed by the HIP dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. | +| GGML_CUDA_MMV_Y | Positive integer | 1 | Block size in y direction for the HIP mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants. | +| GGML_CUDA_KQUANTS_ITER | 1 or 2 | 2 | Number of values processed per iteration and per HIP thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs. | + +### Vulkan + +**With docker**: + +You don't need to install Vulkan SDK. It will be installed inside the container. + +```sh +# Build the image +docker build -t llama-cpp-vulkan -f .devops/llama-cli-vulkan.Dockerfile . + +# Then, use it: +docker run -it --rm -v "$(pwd):/app:Z" --device /dev/dri/renderD128:/dev/dri/renderD128 --device /dev/dri/card1:/dev/dri/card1 llama-cpp-vulkan -m "/app/models/YOUR_MODEL_FILE" -p "Building a website can be done in 10 simple steps:" -n 400 -e -ngl 33 +``` + +**Without docker**: + +Firstly, you need to make sure you have installed [Vulkan SDK](https://vulkan.lunarg.com/doc/view/latest/linux/getting_started_ubuntu.html) + +For example, on Ubuntu 22.04 (jammy), use the command below: + +```bash +wget -qO - https://packages.lunarg.com/lunarg-signing-key-pub.asc | apt-key add - +wget -qO /etc/apt/sources.list.d/lunarg-vulkan-jammy.list https://packages.lunarg.com/vulkan/lunarg-vulkan-jammy.list +apt update -y +apt-get install -y vulkan-sdk +# To verify the installation, use the command below: +vulkaninfo +``` + +Alternatively your package manager might be able to provide the appropriate libraries. +For example for Ubuntu 22.04 you can install `libvulkan-dev` instead. +For Fedora 40, you can install `vulkan-devel`, `glslc` and `glslang` packages. + +Then, build llama.cpp using the cmake command below: + +```bash +cmake -B build -DGGML_VULKAN=1 +cmake --build build --config Release +# Test the output binary (with "-ngl 33" to offload all layers to GPU) +./bin/llama-cli -m "PATH_TO_MODEL" -p "Hi you how are you" -n 50 -e -ngl 33 -t 4 + +# You should see in the output, ggml_vulkan detected your GPU. For example: +# ggml_vulkan: Using Intel(R) Graphics (ADL GT2) | uma: 1 | fp16: 1 | warp size: 32 +``` + +### Android + +To read documentation for how to build on Android, [click here](./android.md) diff --git a/docs/HOWTO-add-model.md b/docs/development/HOWTO-add-model.md similarity index 96% rename from docs/HOWTO-add-model.md rename to docs/development/HOWTO-add-model.md index 3eec077ea..2712b66c1 100644 --- a/docs/HOWTO-add-model.md +++ b/docs/development/HOWTO-add-model.md @@ -1,4 +1,4 @@ -## Add a new model architecture to `llama.cpp` +# Add a new model architecture to `llama.cpp` Adding a model requires few steps: @@ -17,7 +17,7 @@ Also, it is important to check that the examples and main ggml backends (CUDA, M ### 1. Convert the model to GGUF This step is done in python with a `convert` script using the [gguf](https://pypi.org/project/gguf/) library. -Depending on the model architecture, you can use either [convert-hf-to-gguf.py](../convert-hf-to-gguf.py) or [examples/convert-legacy-llama.py](../examples/convert-legacy-llama.py) (for `llama/llama2` models in `.pth` format). +Depending on the model architecture, you can use either [convert_hf_to_gguf.py](../convert_hf_to_gguf.py) or [examples/convert_legacy_llama.py](../examples/convert_legacy_llama.py) (for `llama/llama2` models in `.pth` format). The convert script reads the model configuration, tokenizer, tensor names+data and converts them to GGUF metadata and tensors. diff --git a/docs/debugging-tests.md b/docs/development/debugging-tests.md similarity index 100% rename from docs/debugging-tests.md rename to docs/development/debugging-tests.md diff --git a/docs/llama-star/idea-arch.key b/docs/development/llama-star/idea-arch.key similarity index 100% rename from docs/llama-star/idea-arch.key rename to docs/development/llama-star/idea-arch.key diff --git a/docs/llama-star/idea-arch.pdf b/docs/development/llama-star/idea-arch.pdf similarity index 100% rename from docs/llama-star/idea-arch.pdf rename to docs/development/llama-star/idea-arch.pdf diff --git a/docs/token_generation_performance_tips.md b/docs/development/token_generation_performance_tips.md similarity index 100% rename from docs/token_generation_performance_tips.md rename to docs/development/token_generation_performance_tips.md diff --git a/docs/docker.md b/docs/docker.md new file mode 100644 index 000000000..d8922d77d --- /dev/null +++ b/docs/docker.md @@ -0,0 +1,86 @@ +# Docker + +## Prerequisites +* Docker must be installed and running on your system. +* Create a folder to store big models & intermediate files (ex. /llama/models) + +## Images +We have three Docker images available for this project: + +1. `ghcr.io/ggerganov/llama.cpp:full`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. (platforms: `linux/amd64`, `linux/arm64`) +2. `ghcr.io/ggerganov/llama.cpp:light`: This image only includes the main executable file. (platforms: `linux/amd64`, `linux/arm64`) +3. `ghcr.io/ggerganov/llama.cpp:server`: This image only includes the server executable file. (platforms: `linux/amd64`, `linux/arm64`) + +Additionally, there the following images, similar to the above: + +- `ghcr.io/ggerganov/llama.cpp:full-cuda`: Same as `full` but compiled with CUDA support. (platforms: `linux/amd64`) +- `ghcr.io/ggerganov/llama.cpp:light-cuda`: Same as `light` but compiled with CUDA support. (platforms: `linux/amd64`) +- `ghcr.io/ggerganov/llama.cpp:server-cuda`: Same as `server` but compiled with CUDA support. (platforms: `linux/amd64`) +- `ghcr.io/ggerganov/llama.cpp:full-rocm`: Same as `full` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) +- `ghcr.io/ggerganov/llama.cpp:light-rocm`: Same as `light` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) +- `ghcr.io/ggerganov/llama.cpp:server-rocm`: Same as `server` but compiled with ROCm support. (platforms: `linux/amd64`, `linux/arm64`) + +The GPU enabled images are not currently tested by CI beyond being built. They are not built with any variation from the ones in the Dockerfiles defined in [.devops/](.devops/) and the GitHub Action defined in [.github/workflows/docker.yml](.github/workflows/docker.yml). If you need different settings (for example, a different CUDA or ROCm library, you'll need to build the images locally for now). + +## Usage + +The easiest way to download the models, convert them to ggml and optimize them is with the --all-in-one command which includes the full docker image. + +Replace `/path/to/models` below with the actual path where you downloaded the models. + +```bash +docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --all-in-one "/models/" 7B +``` + +On completion, you are ready to play! + +```bash +docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:full --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 +``` + +or with a light image: + +```bash +docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 +``` + +or with a server image: + +```bash +docker run -v /path/to/models:/models -p 8000:8000 ghcr.io/ggerganov/llama.cpp:server -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 +``` + +## Docker With CUDA + +Assuming one has the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) properly installed on Linux, or is using a GPU enabled cloud, `cuBLAS` should be accessible inside the container. + +## Building Docker locally + +```bash +docker build -t local/llama.cpp:full-cuda -f .devops/full-cuda.Dockerfile . +docker build -t local/llama.cpp:light-cuda -f .devops/llama-cli-cuda.Dockerfile . +docker build -t local/llama.cpp:server-cuda -f .devops/llama-server-cuda.Dockerfile . +``` + +You may want to pass in some different `ARGS`, depending on the CUDA environment supported by your container host, as well as the GPU architecture. + +The defaults are: + +- `CUDA_VERSION` set to `11.7.1` +- `CUDA_DOCKER_ARCH` set to `all` + +The resulting images, are essentially the same as the non-CUDA images: + +1. `local/llama.cpp:full-cuda`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization. +2. `local/llama.cpp:light-cuda`: This image only includes the main executable file. +3. `local/llama.cpp:server-cuda`: This image only includes the server executable file. + +## Usage + +After building locally, Usage is similar to the non-CUDA examples, but you'll need to add the `--gpus` flag. You will also want to use the `--n-gpu-layers` flag. + +```bash +docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1 +docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.gguf -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1 +docker run --gpus all -v /path/to/models:/models local/llama.cpp:server-cuda -m /models/7B/ggml-model-q4_0.gguf --port 8000 --host 0.0.0.0 -n 512 --n-gpu-layers 1 +``` diff --git a/docs/install.md b/docs/install.md new file mode 100644 index 000000000..10a568506 --- /dev/null +++ b/docs/install.md @@ -0,0 +1,39 @@ +# Install pre-built version of llama.cpp + +## Homebrew + +On Mac and Linux, the homebrew package manager can be used via + +```sh +brew install llama.cpp +``` +The formula is automatically updated with new `llama.cpp` releases. More info: https://github.com/ggerganov/llama.cpp/discussions/7668 + +## Nix + +On Mac and Linux, the Nix package manager can be used via + +```sh +nix profile install nixpkgs#llama-cpp +``` +For flake enabled installs. + +Or + +```sh +nix-env --file '' --install --attr llama-cpp +``` + +For non-flake enabled installs. + +This expression is automatically updated within the [nixpkgs repo](https://github.com/NixOS/nixpkgs/blob/nixos-24.05/pkgs/by-name/ll/llama-cpp/package.nix#L164). + +## Flox + +On Mac and Linux, Flox can be used to install llama.cpp within a Flox environment via + +```sh +flox install llama-cpp +``` + +Flox follows the nixpkgs build of llama.cpp. diff --git a/examples/batched.swift/Sources/main.swift b/examples/batched.swift/Sources/main.swift index dbbd06da5..616494d2d 100644 --- a/examples/batched.swift/Sources/main.swift +++ b/examples/batched.swift/Sources/main.swift @@ -229,7 +229,7 @@ private func tokenize(text: String, add_bos: Bool) -> [llama_token] { private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String? { var result = [CChar](repeating: 0, count: 8) - let nTokens = llama_token_to_piece(model, token, &result, Int32(result.count), false) + let nTokens = llama_token_to_piece(model, token, &result, Int32(result.count), 0, false) if nTokens < 0 { let actualTokensCount = -Int(nTokens) result = .init(repeating: 0, count: actualTokensCount) @@ -238,6 +238,7 @@ private func token_to_piece(token: llama_token, buffer: inout [CChar]) -> String token, &result, Int32(result.count), + 0, false ) assert(check == actualTokensCount) diff --git a/examples/batched/batched.cpp b/examples/batched/batched.cpp index 62d9b144d..2442e954d 100644 --- a/examples/batched/batched.cpp +++ b/examples/batched/batched.cpp @@ -93,14 +93,34 @@ int main(int argc, char ** argv) { // create a llama_batch // we use this object to submit token data for decoding - llama_batch batch = llama_batch_init(std::max(tokens_list.size(), (size_t)n_parallel), 0, 1); + llama_batch batch = llama_batch_init(std::max(tokens_list.size(), (size_t) n_parallel), 0, n_parallel); + + std::vector seq_ids(n_parallel, 0); + for (int32_t i = 0; i < n_parallel; ++i) { + seq_ids[i] = i; + } // evaluate the initial prompt for (size_t i = 0; i < tokens_list.size(); ++i) { - llama_batch_add(batch, tokens_list[i], i, { 0 }, false); + llama_batch_add(batch, tokens_list[i], i, seq_ids, false); } GGML_ASSERT(batch.n_tokens == (int) tokens_list.size()); + if (llama_model_has_encoder(model)) { + if (llama_encode(ctx, batch)) { + LOG_TEE("%s : failed to eval\n", __func__); + return 1; + } + + llama_token decoder_start_token_id = llama_model_decoder_start_token(model); + if (decoder_start_token_id == -1) { + decoder_start_token_id = llama_token_bos(model); + } + + llama_batch_clear(batch); + llama_batch_add(batch, decoder_start_token_id, 0, seq_ids, false); + } + // llama_decode will output logits only for the last token of the prompt batch.logits[batch.n_tokens - 1] = true; @@ -109,11 +129,11 @@ int main(int argc, char ** argv) { return 1; } - // assign the system KV cache to all parallel sequences - // this way, the parallel sequences will "reuse" the prompt tokens without having to copy them - for (int32_t i = 1; i < n_parallel; ++i) { - llama_kv_cache_seq_cp(ctx, 0, i, -1, -1); - } + //// assign the system KV cache to all parallel sequences + //// this way, the parallel sequences will "reuse" the prompt tokens without having to copy them + //for (int32_t i = 1; i < n_parallel; ++i) { + // llama_kv_cache_seq_cp(ctx, 0, i, -1, -1); + //} if (n_parallel > 1) { LOG_TEE("\n\n%s: generating %d sequences ...\n", __func__, n_parallel); diff --git a/examples/convert-legacy-llama.py b/examples/convert_legacy_llama.py similarity index 100% rename from examples/convert-legacy-llama.py rename to examples/convert_legacy_llama.py diff --git a/examples/finetune/convert-finetune-checkpoint-to-gguf.py b/examples/finetune/convert_finetune_checkpoint_to_gguf.py similarity index 100% rename from examples/finetune/convert-finetune-checkpoint-to-gguf.py rename to examples/finetune/convert_finetune_checkpoint_to_gguf.py diff --git a/examples/json-schema-pydantic-example.py b/examples/json_schema_pydantic_example.py similarity index 98% rename from examples/json-schema-pydantic-example.py rename to examples/json_schema_pydantic_example.py index 2a24f8118..c7ca7b8d9 100644 --- a/examples/json-schema-pydantic-example.py +++ b/examples/json_schema_pydantic_example.py @@ -1,7 +1,7 @@ # Usage: #! ./llama-server -m some-model.gguf & #! pip install pydantic -#! python json-schema-pydantic-example.py +#! python json_schema_pydantic_example.py from pydantic import BaseModel, Extra, TypeAdapter from annotated_types import MinLen diff --git a/examples/llama.swiftui/llama.cpp.swift/LibLlama.swift b/examples/llama.swiftui/llama.cpp.swift/LibLlama.swift index 737f882fb..2a3f9f758 100644 --- a/examples/llama.swiftui/llama.cpp.swift/LibLlama.swift +++ b/examples/llama.swiftui/llama.cpp.swift/LibLlama.swift @@ -322,7 +322,7 @@ actor LlamaContext { defer { result.deallocate() } - let nTokens = llama_token_to_piece(model, token, result, 8, false) + let nTokens = llama_token_to_piece(model, token, result, 8, 0, false) if nTokens < 0 { let newResult = UnsafeMutablePointer.allocate(capacity: Int(-nTokens)) @@ -330,7 +330,7 @@ actor LlamaContext { defer { newResult.deallocate() } - let nNewTokens = llama_token_to_piece(model, token, newResult, -nTokens, false) + let nNewTokens = llama_token_to_piece(model, token, newResult, -nTokens, 0, false) let bufferPointer = UnsafeBufferPointer(start: newResult, count: Int(nNewTokens)) return Array(bufferPointer) } else { diff --git a/examples/llava/MobileVLM-README.md b/examples/llava/MobileVLM-README.md index f6c619c87..06a65fba4 100644 --- a/examples/llava/MobileVLM-README.md +++ b/examples/llava/MobileVLM-README.md @@ -30,16 +30,16 @@ git clone https://huggingface.co/mtgv/MobileVLM-1.7B git clone https://huggingface.co/openai/clip-vit-large-patch14-336 ``` -2. Use `llava-surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents: +2. Use `llava_surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents: ```sh -python ./examples/llava/llava-surgery.py -m path/to/MobileVLM-1.7B +python ./examples/llava/llava_surgery.py -m path/to/MobileVLM-1.7B ``` -3. Use `convert-image-encoder-to-gguf.py` with `--projector-type ldp` (for **V2** please use `--projector-type ldpv2`) to convert the LLaVA image encoder to GGUF: +3. Use `convert_image_encoder_to_gguf.py` with `--projector-type ldp` (for **V2** please use `--projector-type ldpv2`) to convert the LLaVA image encoder to GGUF: ```sh -python ./examples/llava/convert-image-encoder-to-gguf \ +python ./examples/llava/convert_image_encoder_to_gguf \ -m path/to/clip-vit-large-patch14-336 \ --llava-projector path/to/MobileVLM-1.7B/llava.projector \ --output-dir path/to/MobileVLM-1.7B \ @@ -47,17 +47,17 @@ python ./examples/llava/convert-image-encoder-to-gguf \ ``` ```sh -python ./examples/llava/convert-image-encoder-to-gguf \ +python ./examples/llava/convert_image_encoder_to_gguf \ -m path/to/clip-vit-large-patch14-336 \ --llava-projector path/to/MobileVLM-1.7B_V2/llava.projector \ --output-dir path/to/MobileVLM-1.7B_V2 \ --projector-type ldpv2 ``` -4. Use `examples/convert-legacy-llama.py` to convert the LLaMA part of LLaVA to GGUF: +4. Use `examples/convert_legacy_llama.py` to convert the LLaMA part of LLaVA to GGUF: ```sh -python ./examples/convert-legacy-llama.py path/to/MobileVLM-1.7B +python ./examples/convert_legacy_llama.py path/to/MobileVLM-1.7B ``` 5. Use `quantize` to convert LLaMA part's DataType from `fp16` to `q4_k` diff --git a/examples/llava/README.md b/examples/llava/README.md index f4554de67..012451361 100644 --- a/examples/llava/README.md +++ b/examples/llava/README.md @@ -38,22 +38,22 @@ git clone https://huggingface.co/openai/clip-vit-large-patch14-336 pip install -r examples/llava/requirements.txt ``` -3. Use `llava-surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents: +3. Use `llava_surgery.py` to split the LLaVA model to LLaMA and multimodel projector constituents: ```sh -python ./examples/llava/llava-surgery.py -m ../llava-v1.5-7b +python ./examples/llava/llava_surgery.py -m ../llava-v1.5-7b ``` -4. Use `convert-image-encoder-to-gguf.py` to convert the LLaVA image encoder to GGUF: +4. Use `convert_image_encoder_to_gguf.py` to convert the LLaVA image encoder to GGUF: ```sh -python ./examples/llava/convert-image-encoder-to-gguf.py -m ../clip-vit-large-patch14-336 --llava-projector ../llava-v1.5-7b/llava.projector --output-dir ../llava-v1.5-7b +python ./examples/llava/convert_image_encoder_to_gguf.py -m ../clip-vit-large-patch14-336 --llava-projector ../llava-v1.5-7b/llava.projector --output-dir ../llava-v1.5-7b ``` -5. Use `examples/convert-legacy-llama.py` to convert the LLaMA part of LLaVA to GGUF: +5. Use `examples/convert_legacy_llama.py` to convert the LLaMA part of LLaVA to GGUF: ```sh -python ./examples/convert-legacy-llama.py ../llava-v1.5-7b --skip-unknown +python ./examples/convert_legacy_llama.py ../llava-v1.5-7b --skip-unknown ``` Now both the LLaMA part and the image encoder are in the `llava-v1.5-7b` directory. @@ -70,9 +70,9 @@ git clone https://huggingface.co/liuhaotian/llava-v1.6-vicuna-7b pip install -r examples/llava/requirements.txt ``` -3) Use `llava-surgery-v2.py` which also supports llava-1.5 variants pytorch as well as safetensor models: +3) Use `llava_surgery_v2.py` which also supports llava-1.5 variants pytorch as well as safetensor models: ```console -python examples/llava/llava-surgery-v2.py -C -m ../llava-v1.6-vicuna-7b/ +python examples/llava/llava_surgery_v2.py -C -m ../llava-v1.6-vicuna-7b/ ``` - you will find a llava.projector and a llava.clip file in your model directory @@ -86,13 +86,13 @@ curl -s -q https://huggingface.co/cmp-nct/llava-1.6-gguf/raw/main/config_vit.jso 5) Create the visual gguf model: ```console -python ./examples/llava/convert-image-encoder-to-gguf.py -m vit --llava-projector vit/llava.projector --output-dir vit --clip-model-is-vision +python ./examples/llava/convert_image_encoder_to_gguf.py -m vit --llava-projector vit/llava.projector --output-dir vit --clip-model-is-vision ``` - This is similar to llava-1.5, the difference is that we tell the encoder that we are working with the pure vision model part of CLIP 6) Then convert the model to gguf format: ```console -python ./examples/convert-legacy-llama.py ../llava-v1.6-vicuna-7b/ --skip-unknown +python ./examples/convert_legacy_llama.py ../llava-v1.6-vicuna-7b/ --skip-unknown ``` 7) And finally we can run the llava cli using the 1.6 model version: diff --git a/examples/llava/convert-image-encoder-to-gguf.py b/examples/llava/convert_image_encoder_to_gguf.py similarity index 100% rename from examples/llava/convert-image-encoder-to-gguf.py rename to examples/llava/convert_image_encoder_to_gguf.py diff --git a/examples/llava/llava-surgery.py b/examples/llava/llava_surgery.py similarity index 100% rename from examples/llava/llava-surgery.py rename to examples/llava/llava_surgery.py diff --git a/examples/llava/llava-surgery-v2.py b/examples/llava/llava_surgery_v2.py similarity index 100% rename from examples/llava/llava-surgery-v2.py rename to examples/llava/llava_surgery_v2.py diff --git a/examples/llava/requirements.txt b/examples/llava/requirements.txt index 17cb4d5e5..4713f0a34 100644 --- a/examples/llava/requirements.txt +++ b/examples/llava/requirements.txt @@ -1,3 +1,3 @@ --r ../../requirements/requirements-convert-legacy-llama.txt +-r ../../requirements/requirements-convert_legacy_llama.txt pillow~=10.2.0 -torch~=2.1.1 +torch~=2.2.1 diff --git a/examples/main/main.cpp b/examples/main/main.cpp index d512953b9..4ef55c1e6 100644 --- a/examples/main/main.cpp +++ b/examples/main/main.cpp @@ -37,7 +37,8 @@ static gpt_params * g_params; static std::vector * g_input_tokens; static std::ostringstream * g_output_ss; static std::vector * g_output_tokens; -static bool is_interacting = false; +static bool is_interacting = false; +static bool need_insert_eot = false; static bool file_exists(const std::string & path) { std::ifstream f(path.c_str()); @@ -99,7 +100,8 @@ static void write_logfile( static void sigint_handler(int signo) { if (signo == SIGINT) { if (!is_interacting && g_params->interactive) { - is_interacting = true; + is_interacting = true; + need_insert_eot = true; } else { console::cleanup(); printf("\n"); @@ -224,7 +226,14 @@ int main(int argc, char ** argv) { __func__, n_ctx_train, n_ctx); } - LOG_TEE("%s: chat template example: %s\n", __func__, llama_chat_format_example(model, params.chat_template).c_str()); + // print chat template example in conversation mode + if (params.conversation) { + if (params.enable_chat_template) { + LOG_TEE("%s: chat template example: %s\n", __func__, llama_chat_format_example(model, params.chat_template).c_str()); + } else { + LOG_TEE("%s: in-suffix/prefix is specified, chat template will be disabled\n", __func__); + } + } // print system information { @@ -255,13 +264,15 @@ int main(int argc, char ** argv) { } const bool add_bos = llama_should_add_bos_token(model); - GGML_ASSERT(llama_add_eos_token(model) != 1); + if (!llama_model_has_encoder(model)) { + GGML_ASSERT(llama_add_eos_token(model) != 1); + } LOG("add_bos: %d\n", add_bos); std::vector embd_inp; { - auto prompt = (params.conversation && params.enable_chat_template) + auto prompt = (params.conversation && params.enable_chat_template && !params.prompt.empty()) ? chat_add_and_format(model, chat_msgs, "system", params.prompt) // format the system prompt in conversation mode : params.prompt; if (params.interactive_first || !params.prompt.empty() || session_tokens.empty()) { @@ -517,6 +528,24 @@ int main(int argc, char ** argv) { exit(1); } + if (llama_model_has_encoder(model)) { + int enc_input_size = embd_inp.size(); + llama_token * enc_input_buf = embd_inp.data(); + + if (llama_encode(ctx, llama_batch_get_one(enc_input_buf, enc_input_size, 0, 0))) { + LOG_TEE("%s : failed to eval\n", __func__); + return 1; + } + + llama_token decoder_start_token_id = llama_model_decoder_start_token(model); + if (decoder_start_token_id == -1) { + decoder_start_token_id = llama_token_bos(model); + } + + embd_inp.clear(); + embd_inp.push_back(decoder_start_token_id); + } + while ((n_remain != 0 && !is_antiprompt) || params.interactive) { // predict if (!embd.empty()) { @@ -885,6 +914,13 @@ int main(int argc, char ** argv) { LOG("input tokens: %s\n", LOG_TOKENS_TOSTR_PRETTY(ctx, line_inp).c_str()); + // if user stop generation mid-way, we must add EOT to finish model's last response + if (need_insert_eot && format_chat) { + llama_token eot = llama_token_eot(model); + embd_inp.push_back(eot == -1 ? llama_token_eos(model) : eot); + need_insert_eot = false; + } + embd_inp.insert(embd_inp.end(), line_pfx.begin(), line_pfx.end()); embd_inp.insert(embd_inp.end(), line_inp.begin(), line_inp.end()); embd_inp.insert(embd_inp.end(), line_sfx.begin(), line_sfx.end()); diff --git a/examples/passkey/README.md b/examples/passkey/README.md index a48a6283a..2b8e910f9 100644 --- a/examples/passkey/README.md +++ b/examples/passkey/README.md @@ -1,5 +1,8 @@ # llama.cpp/example/passkey +A passkey retrieval task is an evaluation method used to measure a language +models ability to recall information from long contexts. + See the following PRs for more info: - https://github.com/ggerganov/llama.cpp/pull/3856 diff --git a/examples/perplexity/perplexity.cpp b/examples/perplexity/perplexity.cpp index efde8dfdf..dbe445391 100644 --- a/examples/perplexity/perplexity.cpp +++ b/examples/perplexity/perplexity.cpp @@ -1991,6 +1991,12 @@ int main(int argc, char ** argv) { params.n_batch = std::min(params.n_batch, n_kv); } else { params.n_batch = std::min(params.n_batch, params.n_ctx); + if (params.kl_divergence) { + params.n_parallel = 1; + } else { + // ensure there's at least enough seq_ids for HellaSwag + params.n_parallel = std::max(4, params.n_parallel); + } } if (params.ppl_stride > 0) { @@ -2015,9 +2021,6 @@ int main(int argc, char ** argv) { llama_model * model; llama_context * ctx; - // ensure there's at least enough seq_ids for HellaSwag - params.n_parallel = std::max(4, params.n_parallel); - // load the model and apply lora adapter, if any std::tie(model, ctx) = llama_init_from_gpt_params(params); if (model == NULL) { diff --git a/examples/pydantic-models-to-grammar-examples.py b/examples/pydantic_models_to_grammar_examples.py similarity index 100% rename from examples/pydantic-models-to-grammar-examples.py rename to examples/pydantic_models_to_grammar_examples.py diff --git a/examples/quantize/README.md b/examples/quantize/README.md index b78ece4e7..553c2701b 100644 --- a/examples/quantize/README.md +++ b/examples/quantize/README.md @@ -4,7 +4,89 @@ You can also use the [GGUF-my-repo](https://huggingface.co/spaces/ggml-org/gguf- Note: It is synced from llama.cpp `main` every 6 hours. -## Llama 2 7B +Example usage: + +```bash +# obtain the official LLaMA model weights and place them in ./models +ls ./models +llama-2-7b tokenizer_checklist.chk tokenizer.model +# [Optional] for models using BPE tokenizers +ls ./models + vocab.json +# [Optional] for PyTorch .bin models like Mistral-7B +ls ./models + + +# install Python dependencies +python3 -m pip install -r requirements.txt + +# convert the model to ggml FP16 format +python3 convert_hf_to_gguf.py models/mymodel/ + +# quantize the model to 4-bits (using Q4_K_M method) +./llama-quantize ./models/mymodel/ggml-model-f16.gguf ./models/mymodel/ggml-model-Q4_K_M.gguf Q4_K_M + +# update the gguf filetype to current version if older version is now unsupported +./llama-quantize ./models/mymodel/ggml-model-Q4_K_M.gguf ./models/mymodel/ggml-model-Q4_K_M-v2.gguf COPY +``` + +Run the quantized model: + +```bash +# start inference on a gguf model +./llama-cli -m ./models/mymodel/ggml-model-Q4_K_M.gguf -n 128 +``` + +When running the larger models, make sure you have enough disk space to store all the intermediate files. + +## Memory/Disk Requirements + +As the models are currently fully loaded into memory, you will need adequate disk space to save them and sufficient RAM to load them. At the moment, memory and disk requirements are the same. + +| Model | Original size | Quantized size (Q4_0) | +|------:|--------------:|----------------------:| +| 7B | 13 GB | 3.9 GB | +| 13B | 24 GB | 7.8 GB | +| 30B | 60 GB | 19.5 GB | +| 65B | 120 GB | 38.5 GB | + +## Quantization + +Several quantization methods are supported. They differ in the resulting model disk size and inference speed. + +*(outdated)* + +| Model | Measure | F16 | Q4_0 | Q4_1 | Q5_0 | Q5_1 | Q8_0 | +|------:|--------------|-------:|-------:|-------:|-------:|-------:|-------:| +| 7B | perplexity | 5.9066 | 6.1565 | 6.0912 | 5.9862 | 5.9481 | 5.9070 | +| 7B | file size | 13.0G | 3.5G | 3.9G | 4.3G | 4.7G | 6.7G | +| 7B | ms/tok @ 4th | 127 | 55 | 54 | 76 | 83 | 72 | +| 7B | ms/tok @ 8th | 122 | 43 | 45 | 52 | 56 | 67 | +| 7B | bits/weight | 16.0 | 4.5 | 5.0 | 5.5 | 6.0 | 8.5 | +| 13B | perplexity | 5.2543 | 5.3860 | 5.3608 | 5.2856 | 5.2706 | 5.2548 | +| 13B | file size | 25.0G | 6.8G | 7.6G | 8.3G | 9.1G | 13G | +| 13B | ms/tok @ 4th | - | 103 | 105 | 148 | 160 | 131 | +| 13B | ms/tok @ 8th | - | 73 | 82 | 98 | 105 | 128 | +| 13B | bits/weight | 16.0 | 4.5 | 5.0 | 5.5 | 6.0 | 8.5 | + +- [k-quants](https://github.com/ggerganov/llama.cpp/pull/1684) +- recent k-quants improvements and new i-quants + - [#2707](https://github.com/ggerganov/llama.cpp/pull/2707) + - [#2807](https://github.com/ggerganov/llama.cpp/pull/2807) + - [#4773 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4773) + - [#4856 - 2-bit i-quants (inference)](https://github.com/ggerganov/llama.cpp/pull/4856) + - [#4861 - importance matrix](https://github.com/ggerganov/llama.cpp/pull/4861) + - [#4872 - MoE models](https://github.com/ggerganov/llama.cpp/pull/4872) + - [#4897 - 2-bit quantization](https://github.com/ggerganov/llama.cpp/pull/4897) + - [#4930 - imatrix for all k-quants](https://github.com/ggerganov/llama.cpp/pull/4930) + - [#4951 - imatrix on the GPU](https://github.com/ggerganov/llama.cpp/pull/4957) + - [#4969 - imatrix for legacy quants](https://github.com/ggerganov/llama.cpp/pull/4969) + - [#4996 - k-qunats tuning](https://github.com/ggerganov/llama.cpp/pull/4996) + - [#5060 - Q3_K_XS](https://github.com/ggerganov/llama.cpp/pull/5060) + - [#5196 - 3-bit i-quants](https://github.com/ggerganov/llama.cpp/pull/5196) + - [quantization tuning](https://github.com/ggerganov/llama.cpp/pull/5320), [another one](https://github.com/ggerganov/llama.cpp/pull/5334), and [another one](https://github.com/ggerganov/llama.cpp/pull/5361) + +**Llama 2 7B** | Quantization | Bits per Weight (BPW) | |--------------|-----------------------| @@ -18,7 +100,8 @@ Note: It is synced from llama.cpp `main` every 6 hours. | Q5_K_M | 5.68 | | Q6_K | 6.56 | -## Llama 2 13B +**Llama 2 13B** + Quantization | Bits per Weight (BPW) -- | -- Q2_K | 3.34 @@ -31,7 +114,7 @@ Q5_K_S | 5.51 Q5_K_M | 5.67 Q6_K | 6.56 -# Llama 2 70B +**Llama 2 70B** Quantization | Bits per Weight (BPW) -- | -- diff --git a/examples/regex-to-grammar.py b/examples/regex_to_grammar.py similarity index 100% rename from examples/regex-to-grammar.py rename to examples/regex_to_grammar.py diff --git a/examples/server/README.md b/examples/server/README.md index 4fab006bb..cb45ee06d 100644 --- a/examples/server/README.md +++ b/examples/server/README.md @@ -366,7 +366,8 @@ Notice that each `probs` is an array of length `n_probs`. "assistant_name": "", "user_name": "", "default_generation_settings": { ... }, - "total_slots": 1 + "total_slots": 1, + "chat_template": "" } ``` @@ -374,8 +375,9 @@ Notice that each `probs` is an array of length `n_probs`. - `user_name` - the required anti-prompt to generate the prompt in case you have specified a system prompt for all slots. - `default_generation_settings` - the default generation settings for the `/completion` endpoint, which has the same fields as the `generation_settings` response object from the `/completion` endpoint. - `total_slots` - the total number of slots for process requests (defined by `--parallel` option) +- `chat_template` - the model's original Jinja2 prompt template -- **POST** `/v1/chat/completions`: OpenAI-compatible Chat Completions API. Given a ChatML-formatted json description in `messages`, it returns the predicted completion. Both synchronous and streaming mode are supported, so scripted and interactive applications work fine. While no strong claims of compatibility with OpenAI API spec is being made, in our experience it suffices to support many apps. Only model with [supported chat template](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) can be used optimally with this endpoint. By default, ChatML template will be used. +- **POST** `/v1/chat/completions`: OpenAI-compatible Chat Completions API. Given a ChatML-formatted json description in `messages`, it returns the predicted completion. Both synchronous and streaming mode are supported, so scripted and interactive applications work fine. While no strong claims of compatibility with OpenAI API spec is being made, in our experience it suffices to support many apps. Only models with a [supported chat template](https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template) can be used optimally with this endpoint. By default, the ChatML template will be used. *Options:* diff --git a/examples/server/server.cpp b/examples/server/server.cpp index d7fb61812..47bea1591 100644 --- a/examples/server/server.cpp +++ b/examples/server/server.cpp @@ -2605,7 +2605,7 @@ int main(int argc, char ** argv) { // if a custom chat template is not supplied, we will use the one that comes with the model (if any) if (params.chat_template.empty()) { if (!ctx_server.validate_model_chat_template()) { - LOG_ERROR("The chat template that comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {}); + LOG_WARNING("The chat template that comes with this model is not yet supported, falling back to chatml. This may cause the model to output suboptimal responses", {}); params.chat_template = "chatml"; } } @@ -2967,11 +2967,20 @@ int main(int argc, char ** argv) { }; const auto handle_props = [&ctx_server](const httplib::Request & req, httplib::Response & res) { + std::string template_key = "tokenizer.chat_template", curr_tmpl; + int32_t tlen = llama_model_meta_val_str(ctx_server.model, template_key.c_str(), nullptr, 0); + if (tlen > 0) { + std::vector curr_tmpl_buf(tlen + 1, 0); + if (llama_model_meta_val_str(ctx_server.model, template_key.c_str(), curr_tmpl_buf.data(), curr_tmpl_buf.size()) == tlen) { + curr_tmpl = std::string(curr_tmpl_buf.data(), tlen); + } + } res.set_header("Access-Control-Allow-Origin", req.get_header_value("Origin")); json data = { { "system_prompt", ctx_server.system_prompt.c_str() }, { "default_generation_settings", ctx_server.default_generation_settings_for_props }, - { "total_slots", ctx_server.params.n_parallel } + { "total_slots", ctx_server.params.n_parallel }, + { "chat_template", curr_tmpl.c_str() } }; res.set_content(data.dump(), "application/json; charset=utf-8"); diff --git a/examples/server-embd.py b/examples/server_embd.py similarity index 100% rename from examples/server-embd.py rename to examples/server_embd.py diff --git a/examples/tokenize/tokenize.cpp b/examples/tokenize/tokenize.cpp index 54c9834af..0180c87d8 100644 --- a/examples/tokenize/tokenize.cpp +++ b/examples/tokenize/tokenize.cpp @@ -30,6 +30,7 @@ static void print_usage_information(const char * argv0, FILE * stream) { fprintf(stream, " --stdin read prompt from standard input.\n"); fprintf(stream, " --no-bos do not ever add a BOS token to the prompt, even if normally the model uses a BOS token.\n"); fprintf(stream, " --log-disable disable logs. Makes stderr quiet when loading the model.\n"); + fprintf(stream, " --show-count print the total number of tokens.\n"); } static void llama_log_callback_null(ggml_log_level level, const char * text, void * user_data) { @@ -195,6 +196,7 @@ int main(int raw_argc, char ** raw_argv) { bool printing_ids = false; bool no_bos = false; bool disable_logging = false; + bool show_token_count = false; const char * model_path = NULL; const char * prompt_path = NULL; const char * prompt_arg = NULL; @@ -249,6 +251,9 @@ int main(int raw_argc, char ** raw_argv) { else if (arg == "--log-disable") { disable_logging = true; } + else if (arg == "--show-count") { + show_token_count = true; + } else { fprintf(stderr, "Error: unknown option '%s'\n", argv[iarg].c_str()); return 1; @@ -384,6 +389,9 @@ int main(int raw_argc, char ** raw_argv) { printf("]\n"); } + if (show_token_count) { + printf("Total number of tokens: %ld\n", tokens.size()); + } // silence valgrind llama_free(ctx); llama_free_model(model); diff --git a/examples/train-text-from-scratch/convert-train-checkpoint-to-gguf.py b/examples/train-text-from-scratch/convert_train_checkpoint_to_gguf.py similarity index 100% rename from examples/train-text-from-scratch/convert-train-checkpoint-to-gguf.py rename to examples/train-text-from-scratch/convert_train_checkpoint_to_gguf.py diff --git a/ggml/src/CMakeLists.txt b/ggml/src/CMakeLists.txt index 08b71d410..c6694df67 100644 --- a/ggml/src/CMakeLists.txt +++ b/ggml/src/CMakeLists.txt @@ -490,7 +490,7 @@ if (GGML_SYCL) set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fsycl-targets=nvptx64-nvidia-cuda") add_compile_definitions(GGML_SYCL_WARP_SIZE=32) else() - add_compile_definitions(GGML_SYCL_WARP_SIZE=32) + add_compile_definitions(GGML_SYCL_WARP_SIZE=16) endif() file(GLOB GGML_HEADERS_SYCL "ggml-sycl/*.hpp") @@ -1175,4 +1175,5 @@ endif() if (BUILD_SHARED_LIBS) set_target_properties(ggml PROPERTIES POSITION_INDEPENDENT_CODE ON) + target_compile_definitions(ggml PRIVATE GGML_SHARED GGML_BUILD) endif() diff --git a/ggml/src/ggml-cuda/common.cuh b/ggml/src/ggml-cuda/common.cuh index 472f4ace1..4ff06b871 100644 --- a/ggml/src/ggml-cuda/common.cuh +++ b/ggml/src/ggml-cuda/common.cuh @@ -227,6 +227,10 @@ typedef float2 dfloat2; #define RDNA2 #endif +#if defined(__gfx1010__) || defined(__gfx1012__) +#define RDNA1 +#endif + #ifndef __has_builtin #define __has_builtin(x) 0 #endif diff --git a/ggml/src/ggml-cuda/fattn-common.cuh b/ggml/src/ggml-cuda/fattn-common.cuh index 650780fd2..f24312dd0 100644 --- a/ggml/src/ggml-cuda/fattn-common.cuh +++ b/ggml/src/ggml-cuda/fattn-common.cuh @@ -68,7 +68,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_0( const int iqs4 = k_KQ % QI4_0; const int shift = k_KQ & (QI8_1/2); - const int v = (get_int_from_uint8(K_q4_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; + const int v = (get_int_b2(K_q4_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; const int u = Q_q8[k_KQ_0/WARP_SIZE]; const int sumi = ggml_cuda_dp4a(v, u, 0); @@ -108,7 +108,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q4_1( const int iqs4 = k_KQ % QI4_1; const int shift = k_KQ & (QI8_1/2); - const int v = (get_int_from_uint8_aligned(K_q4_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; + const int v = (get_int_b4(K_q4_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; const int u = Q_q8[k_KQ_0/WARP_SIZE]; const int sumi = ggml_cuda_dp4a(v, u, 0); @@ -153,8 +153,8 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_0( const int iqs8 = k_KQ % QI8_1; const int shift = k_KQ & (QI8_1/2); - int v = (get_int_from_uint8(K_q5_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; - const int vh = get_int_from_uint8(K_q5_0[ib].qh, 0) >> (iqs8 * QI5_0); + int v = (get_int_b2(K_q5_0[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; + const int vh = get_int_b2(K_q5_0[ib].qh, 0) >> (iqs8 * QI5_0); v |= (vh << 4) & 0x00000010; // 0 -> 4 v |= (vh << 11) & 0x00001000; // 1 -> 12 v |= (vh << 18) & 0x00100000; // 2 -> 20 @@ -200,8 +200,8 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q5_1( const int iqs8 = k_KQ % QI8_1; const int shift = k_KQ & (QI8_1/2); - int v = (get_int_from_uint8(K_q5_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; - const int vh = get_int_from_uint8(K_q5_1[ib].qh, 0) >> (iqs8 * QI5_1); + int v = (get_int_b2(K_q5_1[ib].qs, iqs4) >> shift) & 0x0F0F0F0F; + const int vh = get_int_b2(K_q5_1[ib].qh, 0) >> (iqs8 * QI5_1); v |= (vh << 4) & 0x00000010; // 0 -> 4 v |= (vh << 11) & 0x00001000; // 1 -> 12 v |= (vh << 18) & 0x00100000; // 2 -> 20 @@ -249,7 +249,7 @@ static __device__ __forceinline__ T vec_dot_fattn_vec_KQ_q8_0( const int ib = k_KQ / QI8_0; const int iqs = k_KQ % QI8_0; - const int v = get_int_from_int8(K_q8_0[ib].qs, iqs); + const int v = get_int_b2(K_q8_0[ib].qs, iqs); T Q_d; if (std::is_same::value) { @@ -408,7 +408,7 @@ static __device__ __forceinline__ T dequantize_1_q5_0(const void * __restrict__ const T d = x[ib].d; const int ql0 = x[ib].qs[iqs]; - const int qh0 = get_int_from_uint8(x[ib].qh, 0); + const int qh0 = get_int_b2(x[ib].qh, 0); const int ql = ((ql0 >> (4*shift)) & 0x0F); const int qh = ((qh0 >> idq) << 4) & 0x10; const int q = (ql | qh) - 16; @@ -433,7 +433,7 @@ static __device__ __forceinline__ T dequantize_1_q5_1(const void * __restrict__ const half2 dm = x[ib].dm; const int ql0 = x[ib].qs[iqs]; - const int qh0 = get_int_from_uint8_aligned(x[ib].qh, 0); + const int qh0 = get_int_b4(x[ib].qh, 0); const int ql = ((ql0 >> (4*shift)) & 0x0F); const int qh = ((qh0 >> idq) << 4) & 0x10; const int q = (ql | qh); diff --git a/ggml/src/ggml-cuda/mmq.cu b/ggml/src/ggml-cuda/mmq.cu index 0308beacc..3af2eec2f 100644 --- a/ggml/src/ggml-cuda/mmq.cu +++ b/ggml/src/ggml-cuda/mmq.cu @@ -59,6 +59,12 @@ void ggml_cuda_op_mul_mat_q( case GGML_TYPE_Q6_K: mul_mat_q_case(ctx, args, stream); break; + case GGML_TYPE_IQ4_XS: + mul_mat_q_case(ctx, args, stream); + break; + case GGML_TYPE_IQ4_NL: + mul_mat_q_case(ctx, args, stream); + break; default: GGML_ASSERT(false); break; @@ -87,6 +93,8 @@ bool ggml_cuda_should_use_mmq(enum ggml_type type, int cc, int64_t ne11) { case GGML_TYPE_Q4_K: case GGML_TYPE_Q5_K: case GGML_TYPE_Q6_K: + case GGML_TYPE_IQ4_XS: + case GGML_TYPE_IQ4_NL: mmq_supported = true; break; default: diff --git a/ggml/src/ggml-cuda/mmq.cuh b/ggml/src/ggml-cuda/mmq.cuh index 1396e7a75..118e34d28 100644 --- a/ggml/src/ggml-cuda/mmq.cuh +++ b/ggml/src/ggml-cuda/mmq.cuh @@ -60,12 +60,16 @@ static constexpr __device__ int get_mmq_x_max_device() { } static constexpr int get_mmq_y_host(const int cc) { - return int8_mma_available(cc) || cc >= CC_VOLTA ? 128 : 64; + return cc >= CC_OFFSET_AMD ? (cc == CC_RDNA1 ? 64 : 128) : (cc >= CC_VOLTA ? 128 : 64); } static constexpr __device__ int get_mmq_y_device() { #if defined(GGML_USE_HIPBLAS) && defined(__HIP_PLATFORM_AMD__) +#if defined(RDNA1) + return 64; +#else return 128; +#endif // defined RDNA1 #else #if __CUDA_ARCH__ >= CC_VOLTA return 128; @@ -88,15 +92,17 @@ static constexpr __device__ int get_mmq_y_device() { static constexpr __host__ __device__ tile_x_sizes mmq_get_dp4a_tile_x_sizes(ggml_type type, int mmq_y) { return type == GGML_TYPE_Q4_0 ? MMQ_DP4A_TXS_Q4_0 : - type == GGML_TYPE_Q4_1 ? MMQ_DP4A_TXS_Q4_1 : - type == GGML_TYPE_Q5_0 ? MMQ_DP4A_TXS_Q5_0 : - type == GGML_TYPE_Q5_1 ? MMQ_DP4A_TXS_Q5_1 : - type == GGML_TYPE_Q8_0 ? MMQ_DP4A_TXS_Q8_0 : - type == GGML_TYPE_Q2_K ? MMQ_DP4A_TXS_Q2_K : - type == GGML_TYPE_Q3_K ? MMQ_DP4A_TXS_Q3_K : - type == GGML_TYPE_Q4_K ? MMQ_DP4A_TXS_Q4_K : - type == GGML_TYPE_Q5_K ? MMQ_DP4A_TXS_Q5_K : - type == GGML_TYPE_Q6_K ? MMQ_DP4A_TXS_Q6_K : + type == GGML_TYPE_Q4_1 ? MMQ_DP4A_TXS_Q4_1 : + type == GGML_TYPE_Q5_0 ? MMQ_DP4A_TXS_Q5_0 : + type == GGML_TYPE_Q5_1 ? MMQ_DP4A_TXS_Q5_1 : + type == GGML_TYPE_Q8_0 ? MMQ_DP4A_TXS_Q8_0 : + type == GGML_TYPE_Q2_K ? MMQ_DP4A_TXS_Q2_K : + type == GGML_TYPE_Q3_K ? MMQ_DP4A_TXS_Q3_K : + type == GGML_TYPE_Q4_K ? MMQ_DP4A_TXS_Q4_K : + type == GGML_TYPE_Q5_K ? MMQ_DP4A_TXS_Q5_K : + type == GGML_TYPE_Q6_K ? MMQ_DP4A_TXS_Q6_K : + type == GGML_TYPE_IQ4_XS ? MMQ_DP4A_TXS_Q5_0 : + type == GGML_TYPE_IQ4_NL ? MMQ_DP4A_TXS_Q5_0 : tile_x_sizes{0, 0, 0}; } @@ -124,15 +130,17 @@ static_assert(MMQ_MMA_TILE_X_K_Q6_K % 8 == 4, "Wrong padding."); static constexpr __host__ __device__ int mmq_get_mma_tile_x_k(ggml_type type) { return type == GGML_TYPE_Q4_0 ? MMQ_MMA_TILE_X_K_Q4_0 : - type == GGML_TYPE_Q4_1 ? MMQ_MMA_TILE_X_K_Q4_1 : - type == GGML_TYPE_Q5_0 ? MMQ_MMA_TILE_X_K_Q5_0 : - type == GGML_TYPE_Q5_1 ? MMQ_MMA_TILE_X_K_Q5_1 : - type == GGML_TYPE_Q8_0 ? MMQ_MMA_TILE_X_K_Q8_0 : - type == GGML_TYPE_Q2_K ? MMQ_MMA_TILE_X_K_Q2_K : - type == GGML_TYPE_Q3_K ? MMQ_MMA_TILE_X_K_Q3_K : - type == GGML_TYPE_Q4_K ? MMQ_MMA_TILE_X_K_Q4_K : - type == GGML_TYPE_Q5_K ? MMQ_MMA_TILE_X_K_Q5_K : - type == GGML_TYPE_Q6_K ? MMQ_MMA_TILE_X_K_Q6_K : + type == GGML_TYPE_Q4_1 ? MMQ_MMA_TILE_X_K_Q4_1 : + type == GGML_TYPE_Q5_0 ? MMQ_MMA_TILE_X_K_Q5_0 : + type == GGML_TYPE_Q5_1 ? MMQ_MMA_TILE_X_K_Q5_1 : + type == GGML_TYPE_Q8_0 ? MMQ_MMA_TILE_X_K_Q8_0 : + type == GGML_TYPE_Q2_K ? MMQ_MMA_TILE_X_K_Q2_K : + type == GGML_TYPE_Q3_K ? MMQ_MMA_TILE_X_K_Q3_K : + type == GGML_TYPE_Q4_K ? MMQ_MMA_TILE_X_K_Q4_K : + type == GGML_TYPE_Q5_K ? MMQ_MMA_TILE_X_K_Q5_K : + type == GGML_TYPE_Q6_K ? MMQ_MMA_TILE_X_K_Q6_K : + type == GGML_TYPE_IQ4_XS ? MMQ_MMA_TILE_X_K_Q5_0 : + type == GGML_TYPE_IQ4_NL ? MMQ_MMA_TILE_X_K_Q5_0 : 0; } @@ -181,9 +189,9 @@ template static __device__ __forceinlin const block_q4_0 * bxi = (const block_q4_0 *) x + kbx0 + i*stride + kbx; #ifdef INT8_MMA_AVAILABLE - x_qs[i*MMQ_MMA_TILE_X_K_Q4_0 + threadIdx.x] = get_int_from_uint8(bxi->qs, kqsx); + x_qs[i*MMQ_MMA_TILE_X_K_Q4_0 + threadIdx.x] = get_int_b2(bxi->qs, kqsx); #else - x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_from_uint8(bxi->qs, kqsx); + x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_b2(bxi->qs, kqsx); #endif // INT8_MMA_AVAILABLE } @@ -344,9 +352,9 @@ template static __device__ __forceinlin const block_q4_1 * bxi = (const block_q4_1 *) x + kbx0 + i*stride + kbx; #ifdef INT8_MMA_AVAILABLE - x_qs[i*MMQ_MMA_TILE_X_K_Q4_1 + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx); + x_qs[i*MMQ_MMA_TILE_X_K_Q4_1 + threadIdx.x] = get_int_b4(bxi->qs, kqsx); #else - x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx); + x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_b4(bxi->qs, kqsx); #endif // INT8_MMA_AVAILABLE } @@ -505,8 +513,8 @@ template static __device__ __forceinlin const block_q5_0 * bxi = (const block_q5_0 *) x + kbx0 + i*stride + kbx; - const int ql = get_int_from_uint8(bxi->qs, kqsx); - const int qh = get_int_from_uint8(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_0)); + const int ql = get_int_b2(bxi->qs, kqsx); + const int qh = get_int_b2(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_0)); int qs0 = (ql >> 0) & 0x0F0F0F0F; qs0 |= (qh << 4) & 0x00000010; // 0 -> 4 @@ -670,8 +678,8 @@ template static __device__ __forceinlin const block_q5_1 * bxi = (const block_q5_1 *) x + kbx0 + i*stride + kbx; - const int ql = get_int_from_uint8_aligned(bxi->qs, kqsx); - const int qh = get_int_from_uint8_aligned(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_1)); + const int ql = get_int_b4(bxi->qs, kqsx); + const int qh = get_int_b4(bxi->qh, 0) >> (4 * (threadIdx.x % QI5_1)); int qs0 = (ql >> 0) & 0x0F0F0F0F; qs0 |= (qh << 4) & 0x00000010; // 0 -> 4 @@ -835,9 +843,9 @@ template static __device__ __forceinlin const block_q8_0 * bxi = (const block_q8_0 *) x + kbx0 + i*stride + kbx; #ifdef INT8_MMA_AVAILABLE - x_qs[i*MMQ_MMA_TILE_X_K_Q8_0 + threadIdx.x] = get_int_from_int8(bxi->qs, kqsx); + x_qs[i*MMQ_MMA_TILE_X_K_Q8_0 + threadIdx.x] = get_int_b2(bxi->qs, kqsx); #else - x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_from_int8(bxi->qs, kqsx); + x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_b2(bxi->qs, kqsx); #endif // INT8_MMA_AVAILABLE } @@ -980,7 +988,7 @@ template static __device__ __forceinlin const block_q2_K * bxi = (const block_q2_K *) x + kbx0 + i*stride + kbx; - const int x_ql_0 = get_int_from_uint8(bxi->qs, kqsx); + const int x_ql_0 = get_int_b2(bxi->qs, kqsx); #pragma unroll for (int l = 0; l < QR2_K; ++l) { @@ -1162,8 +1170,8 @@ template static __device__ __forceinlin const block_q3_K * bxi = (const block_q3_K *) x + kbx0 + i*stride + kbx; - const int x_ql_0 = get_int_from_uint8(bxi->qs, kqsx); - const int x_qh_0 = get_int_from_uint8(bxi->hmask, kqsx % (QI3_K/2)) >> (4 * (kqsx / (QI3_K/2))); + const int x_ql_0 = get_int_b2(bxi->qs, kqsx); + const int x_qh_0 = get_int_b2(bxi->hmask, kqsx % (QI3_K/2)) >> (4 * (kqsx / (QI3_K/2))); #pragma unroll for (int l = 0; l < QR3_K; ++l) { @@ -1221,11 +1229,11 @@ template static __device__ __forceinlin const int ksc_low = ksc % (QI3_K/8); const int shift_low = 4 * (ksc / (QI3_K/8)); - const int sc_low = (get_int_from_uint8(bxi->scales, ksc_low) >> shift_low) & 0x0F0F0F0F; + const int sc_low = (get_int_b2(bxi->scales, ksc_low) >> shift_low) & 0x0F0F0F0F; const int ksc_high = QI3_K/8; const int shift_high = 2 * ksc; - const int sc_high = ((get_int_from_uint8(bxi->scales, ksc_high) >> shift_high) << 4) & 0x30303030; + const int sc_high = ((get_int_b2(bxi->scales, ksc_high) >> shift_high) << 4) & 0x30303030; const int sc = __vsubss4(sc_low | sc_high, 0x20202020); @@ -1389,9 +1397,9 @@ template static __device__ __forceinlin const block_q4_K * bxi = (const block_q4_K *) x + kbx0 + i*stride + kbx; #ifdef INT8_MMA_AVAILABLE - x_qs[i*MMQ_MMA_TILE_X_K_Q4_K + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx); + x_qs[i*MMQ_MMA_TILE_X_K_Q4_K + threadIdx.x] = get_int_b4(bxi->qs, kqsx); #else - x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_from_uint8_aligned(bxi->qs, kqsx); + x_qs[i*(WARP_SIZE + 1) + threadIdx.x] = get_int_b4(bxi->qs, kqsx); #endif // INT8_MMA_AVAILABLE } @@ -1606,11 +1614,11 @@ template static __device__ __forceinlin const block_q5_K * bxi = (const block_q5_K *) x + kbx0 + i*stride + kbx; const int ky = QR5_K*kqsx; - const int ql = get_int_from_uint8_aligned(bxi->qs, kqsx); + const int ql = get_int_b4(bxi->qs, kqsx); const int ql0 = (ql >> 0) & 0x0F0F0F0F; const int ql1 = (ql >> 4) & 0x0F0F0F0F; - const int qh = get_int_from_uint8_aligned(bxi->qh, kqsx % (QI5_K/4)); + const int qh = get_int_b4(bxi->qh, kqsx % (QI5_K/4)); const int qh0 = ((qh >> (2 * (kqsx / (QI5_K/4)) + 0)) << 4) & 0x10101010; const int qh1 = ((qh >> (2 * (kqsx / (QI5_K/4)) + 1)) << 4) & 0x10101010; @@ -1828,11 +1836,11 @@ template static __device__ __forceinlin const block_q6_K * bxi = (const block_q6_K *) x + kbx0 + i*stride + kbx; const int ky = QR6_K*kqsx; - const int ql = get_int_from_uint8(bxi->ql, kqsx); + const int ql = get_int_b2(bxi->ql, kqsx); const int ql0 = (ql >> 0) & 0x0F0F0F0F; const int ql1 = (ql >> 4) & 0x0F0F0F0F; - const int qh = get_int_from_uint8(bxi->qh, (QI6_K/4) * (kqsx / (QI6_K/2)) + kqsx % (QI6_K/4)); + const int qh = get_int_b2(bxi->qh, (QI6_K/4) * (kqsx / (QI6_K/2)) + kqsx % (QI6_K/4)); const int qh0 = ((qh >> (2 * ((kqsx % (QI6_K/2)) / (QI6_K/4)))) << 4) & 0x30303030; const int qh1 = (qh >> (2 * ((kqsx % (QI6_K/2)) / (QI6_K/4)))) & 0x30303030; @@ -1879,9 +1887,9 @@ template static __device__ __forceinlin const block_q6_K * bxi = (const block_q6_K *) x + kbx0 + i*stride + (threadIdx.x % (WARP_SIZE/8)) / 4; #ifdef INT8_MMA_AVAILABLE - x_sc[i*MMQ_MMA_TILE_X_K_Q6_K + threadIdx.x % (WARP_SIZE/8)] = get_int_from_int8(bxi->scales, threadIdx.x % (QI6_K/8)); + x_sc[i*MMQ_MMA_TILE_X_K_Q6_K + threadIdx.x % (WARP_SIZE/8)] = get_int_b2(bxi->scales, threadIdx.x % (QI6_K/8)); #else - x_sc[i*(WARP_SIZE/8) + i/8 + threadIdx.x % (WARP_SIZE/8)] = get_int_from_int8(bxi->scales, threadIdx.x % (QI6_K/8)); + x_sc[i*(WARP_SIZE/8) + i/8 + threadIdx.x % (WARP_SIZE/8)] = get_int_b2(bxi->scales, threadIdx.x % (QI6_K/8)); #endif // INT8_MMA_AVAILABLE } } @@ -2014,6 +2022,124 @@ static __device__ __forceinline__ void vec_dot_q6_K_q8_1_mma( #endif // INT8_MMA_AVAILABLE } +template static __device__ __forceinline__ void load_tiles_iq4_nl( + const char * __restrict__ x, int * __restrict__ x_tile, const int & kbx0, const int & i_max, const int & stride) { + +#ifdef INT8_MMA_AVAILABLE + int * x_qs = (int *) x_tile; + float * x_df = (float *) (x_qs + WARP_SIZE*2); +#else + constexpr tile_x_sizes txs = mmq_get_dp4a_tile_x_sizes(GGML_TYPE_IQ4_NL, mmq_y); + int * x_qs = (int *) x_tile; + float * x_df = (float *) (x_qs + txs.qs); +#endif // INT8_MMA_AVAILABLE + + const int kbx = threadIdx.x / QI4_NL; + const int kqsx = threadIdx.x % QI4_NL; + +#pragma unroll + for (int i0 = 0; i0 < mmq_y; i0 += nwarps) { + int i = i0 + threadIdx.y; + + if (need_check) { + i = min(i, i_max); + } + + const block_iq4_nl * bxi = (const block_iq4_nl *) x + kbx0 + i*stride + kbx; + + const int aux_q4 = get_int_b2(bxi->qs, kqsx); + const int2 v = get_int_from_table_16(aux_q4); + const int k0 = 8 * (threadIdx.x / 4) + threadIdx.x % 4; +#ifdef INT8_MMA_AVAILABLE + x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 0] = v.x; + x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 4] = v.y; +#else + x_qs[i*(2*WARP_SIZE + 1) + k0 + 0] = v.x; + x_qs[i*(2*WARP_SIZE + 1) + k0 + 4] = v.y; +#endif // INT8_MMA_AVAILABLE + } + + const int blocks_per_tile_x_row = WARP_SIZE / QI4_NL; + const int kbxd = threadIdx.x % blocks_per_tile_x_row; + +#pragma unroll + for (int i0 = 0; i0 < mmq_y; i0 += nwarps * QI4_NL) { + int i = i0 + threadIdx.y * QI4_NL + threadIdx.x / blocks_per_tile_x_row; + + if (need_check) { + i = min(i, i_max); + } + + const block_iq4_nl * bxi = (const block_iq4_nl *) x + kbx0 + i*stride + kbxd; + +#ifdef INT8_MMA_AVAILABLE + x_df[i*MMQ_MMA_TILE_X_K_Q5_0 + kbxd] = __half2float(bxi->d); +#else + x_df[i*(WARP_SIZE/4) + i/4 + kbxd] = __half2float(bxi->d); +#endif // INT8_MMA_AVAILABLE + } +} + +template static __device__ __forceinline__ void load_tiles_iq4_xs( + const char * __restrict__ x, int * __restrict__ x_tile, const int & kbx0, const int & i_max, const int & stride) { + +#ifdef INT8_MMA_AVAILABLE + int * x_qs = (int *) x_tile; + float * x_df = (float *) (x_qs + WARP_SIZE*2); +#else + constexpr tile_x_sizes txs = mmq_get_dp4a_tile_x_sizes(GGML_TYPE_IQ4_XS, mmq_y); + int * x_qs = (int *) x_tile; + float * x_df = (float *) (x_qs + txs.qs); +#endif // INT8_MMA_AVAILABLE + + const int kbx = 0; // threadIdx.x / QI4_XS + const int kqsx = threadIdx.x; // threadIdx.x % QI4_XS + +#pragma unroll + for (int i0 = 0; i0 < mmq_y; i0 += nwarps) { + int i = i0 + threadIdx.y; + + if (need_check) { + i = min(i, i_max); + } + + const block_iq4_xs * bxi = (const block_iq4_xs *) x + kbx0 + i*stride + kbx; + + const int aux_q4 = get_int_b4(bxi->qs, kqsx); + const int2 v = get_int_from_table_16(aux_q4); + const int k0 = 8 * (threadIdx.x / 4) + threadIdx.x % 4; +#ifdef INT8_MMA_AVAILABLE + x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 0] = v.x; + x_qs[i*MMQ_MMA_TILE_X_K_Q5_0 + k0 + 4] = v.y; +#else + x_qs[i*(2*WARP_SIZE + 1) + k0 + 0] = v.x; + x_qs[i*(2*WARP_SIZE + 1) + k0 + 4] = v.y; +#endif // INT8_MMA_AVAILABLE + } + +#pragma unroll + for (int i0 = 0; i0 < mmq_y; i0 += nwarps * 4) { + int i = i0 + threadIdx.y * 4 + threadIdx.x / (WARP_SIZE/4); + + if (need_check) { + i = min(i, i_max); + } + + const block_iq4_xs * bxi = (const block_iq4_xs *) x + kbx0 + i*stride; + + const float d = __half2float(bxi->d); + + const int ls = ((bxi->scales_l[(threadIdx.x % 8)/2] >> (4*(threadIdx.x % 2))) & 0x0F) + | (((bxi->scales_h >> (2*(threadIdx.x % 8))) & 0x03) << 4); + +#ifdef INT8_MMA_AVAILABLE + x_df[i*MMQ_MMA_TILE_X_K_Q5_0 + threadIdx.x % 8] = d * (ls - 32); +#else + x_df[i*(WARP_SIZE/4) + i/4 + threadIdx.x % 8] = d * (ls - 32); +#endif // INT8_MMA_AVAILABLE + } +} + template static __device__ __forceinline__ void mmq_write_back_dp4a( const float * __restrict__ sum, float * __restrict__ dst, const int & stride, const int & i_max, const int & j_max) { @@ -2163,6 +2289,22 @@ struct mmq_type_traits { static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q6_K_q8_1_dp4a; }; +template +struct mmq_type_traits { + static constexpr int vdr = VDR_IQ4_NL_Q8_1_MMQ; + static constexpr load_tiles_mmq_t load_tiles = load_tiles_iq4_nl; + static constexpr vec_dot_mmq_t vec_dot_mma = vec_dot_q5_0_q8_1_mma; + static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a; +}; + +template +struct mmq_type_traits { + static constexpr int vdr = VDR_IQ4_XS_Q8_1_MMQ; + static constexpr load_tiles_mmq_t load_tiles = load_tiles_iq4_xs; + static constexpr vec_dot_mmq_t vec_dot_mma = vec_dot_q5_0_q8_1_mma; + static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a; +}; + static bool mmq_need_sum(const ggml_type type_x) { switch (type_x) { case GGML_TYPE_Q4_0: @@ -2180,6 +2322,8 @@ static bool mmq_need_sum(const ggml_type type_x) { case GGML_TYPE_Q5_K: return true; case GGML_TYPE_Q6_K: + case GGML_TYPE_IQ4_XS: + case GGML_TYPE_IQ4_NL: return false; default: GGML_ASSERT(false); @@ -2301,8 +2445,11 @@ static __global__ void mul_mat_q( const int nty = (ne01 + mmq_y - 1) / mmq_y; // Number of tiles y // kbc == k block continuous, current index in continuous ijk space. - int64_t kbc = GGML_PAD((int64_t) blockIdx.x *blocks_per_ne00*ntx*nty / gridDim.x, blocks_per_warp); - const int64_t kbc_stop = GGML_PAD((int64_t)(blockIdx.x + 1)*blocks_per_ne00*ntx*nty / gridDim.x, blocks_per_warp); + int64_t kbc = (int64_t) blockIdx.x *blocks_per_ne00*ntx*nty / gridDim.x; + int64_t kbc_stop = (int64_t)(blockIdx.x + 1)*blocks_per_ne00*ntx*nty / gridDim.x; + + kbc -= (kbc % blocks_per_ne00) % blocks_per_warp; + kbc_stop -= (kbc_stop % blocks_per_ne00) % blocks_per_warp; // kb0 == k index when doing the matrix multiplication for an output tile. int kb0_start = kbc % blocks_per_ne00; @@ -2358,8 +2505,11 @@ static __global__ void mul_mat_q_stream_k_fixup( const int bidx_stop = (blockIdx.y*nty + blockIdx.x + 1) * block_num_mmq / (gridDim.y*gridDim.x) + 1; for (int bidx = bidx_start; bidx < bidx_stop; ++bidx) { - const int64_t kbc = GGML_PAD((int64_t) bidx *blocks_per_ne00*ntx*nty / block_num_mmq, blocks_per_warp); - const int64_t kbc_stop = GGML_PAD((int64_t)(bidx + 1)*blocks_per_ne00*ntx*nty / block_num_mmq, blocks_per_warp); + int64_t kbc = (int64_t) bidx *blocks_per_ne00*ntx*nty / block_num_mmq; + int64_t kbc_stop = (int64_t)(bidx + 1)*blocks_per_ne00*ntx*nty / block_num_mmq; + + kbc -= (kbc % blocks_per_ne00) % blocks_per_warp; + kbc_stop -= (kbc_stop % blocks_per_ne00) % blocks_per_warp; // Skip fixup tile if the MMQ CUDA block never wrote anything to it: if (kbc == kbc_stop || kbc_stop % blocks_per_ne00 == 0) { @@ -2598,6 +2748,8 @@ extern DECL_MMQ_CASE(GGML_TYPE_Q3_K); extern DECL_MMQ_CASE(GGML_TYPE_Q4_K); extern DECL_MMQ_CASE(GGML_TYPE_Q5_K); extern DECL_MMQ_CASE(GGML_TYPE_Q6_K); +extern DECL_MMQ_CASE(GGML_TYPE_IQ4_NL); +extern DECL_MMQ_CASE(GGML_TYPE_IQ4_XS); // ------------------------------------------------------------------------------------------------------------------------- diff --git a/ggml/src/ggml-cuda/template-instances/generate_cu_files.py b/ggml/src/ggml-cuda/template-instances/generate_cu_files.py index ea58d0968..ffeb3c27d 100755 --- a/ggml/src/ggml-cuda/template-instances/generate_cu_files.py +++ b/ggml/src/ggml-cuda/template-instances/generate_cu_files.py @@ -22,7 +22,8 @@ SOURCE_FATTN_WMMA_CASE = "DECL_FATTN_WMMA_F16_CASE({head_size}, {cols_per_block} TYPES_MMQ = [ "GGML_TYPE_Q4_0", "GGML_TYPE_Q4_1", "GGML_TYPE_Q5_0", "GGML_TYPE_Q5_1", "GGML_TYPE_Q8_0", - "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K" + "GGML_TYPE_Q2_K", "GGML_TYPE_Q3_K", "GGML_TYPE_Q4_K", "GGML_TYPE_Q5_K", "GGML_TYPE_Q6_K", + "GGML_TYPE_IQ4_NL", "GGML_TYPE_IQ4_XS" ] SOURCE_MMQ = """// This file has been autogenerated by generate_cu_files.py, do not edit manually. diff --git a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu new file mode 100644 index 000000000..eb02fab00 --- /dev/null +++ b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ4_NL); diff --git a/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu new file mode 100644 index 000000000..1eb3b7430 --- /dev/null +++ b/ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu @@ -0,0 +1,5 @@ +// This file has been autogenerated by generate_cu_files.py, do not edit manually. + +#include "../mmq.cuh" + +DECL_MMQ_CASE(GGML_TYPE_IQ4_XS); diff --git a/ggml/src/ggml-cuda/vecdotq.cuh b/ggml/src/ggml-cuda/vecdotq.cuh index 3f3a87c75..1d510484a 100644 --- a/ggml/src/ggml-cuda/vecdotq.cuh +++ b/ggml/src/ggml-cuda/vecdotq.cuh @@ -1,36 +1,8 @@ #include "common.cuh" #include -static __device__ __forceinline__ int get_int_from_int8(const int8_t * x8, const int & i32) { - const uint16_t * x16 = (const uint16_t *) (x8 + sizeof(int) * i32); // assume at least 2 byte alignment - - int x32 = 0; - x32 |= x16[0] << 0; - x32 |= x16[1] << 16; - - return x32; -} - -static __device__ __forceinline__ int get_int_from_uint8(const uint8_t * x8, const int & i32) { - const uint16_t * x16 = (const uint16_t *) (x8 + sizeof(int) * i32); // assume at least 2 byte alignment - - int x32 = 0; - x32 |= x16[0] << 0; - x32 |= x16[1] << 16; - - return x32; -} - -static __device__ __forceinline__ int get_int_from_int8_aligned(const int8_t * x8, const int & i32) { - return *((const int *) (x8 + sizeof(int) * i32)); // assume at least 4 byte alignment -} - -static __device__ __forceinline__ int get_int_from_uint8_aligned(const uint8_t * x8, const int & i32) { - return *((const int *) (x8 + sizeof(int) * i32)); // assume at least 4 byte alignment -} - static __device__ __forceinline__ int get_int_b2(const void * x, const int & i32) { - const uint16_t * x16 = (const uint16_t *) x; + const uint16_t * x16 = (const uint16_t *) x; // assume at least 2 byte alignment int x32 = x16[2*i32 + 0] << 0; x32 |= x16[2*i32 + 1] << 16; @@ -768,6 +740,7 @@ static __device__ __forceinline__ float vec_dot_q6_K_q8_1( } #define VDR_IQ2_XXS_Q8_1_MMVQ 2 +#define VDR_IQ2_XXS_Q8_1_MMQ 2 static __device__ __forceinline__ float vec_dot_iq2_xxs_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -802,6 +775,7 @@ static __device__ __forceinline__ float vec_dot_iq2_xxs_q8_1( } #define VDR_IQ2_XS_Q8_1_MMVQ 2 +#define VDR_IQ2_XS_Q8_1_MMQ 2 static __device__ __forceinline__ float vec_dot_iq2_xs_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -840,6 +814,7 @@ static __device__ __forceinline__ float vec_dot_iq2_xs_q8_1( } #define VDR_IQ2_S_Q8_1_MMVQ 2 +#define VDR_IQ2_S_Q8_1_MMQ 2 static __device__ __forceinline__ float vec_dot_iq2_s_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -887,6 +862,7 @@ static __device__ __forceinline__ float vec_dot_iq2_s_q8_1( } #define VDR_IQ3_XXS_Q8_1_MMVQ 2 +#define VDR_IQ3_XXS_Q8_1_MMQ 2 static __device__ __forceinline__ float vec_dot_iq3_xxs_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -921,6 +897,7 @@ static __device__ __forceinline__ float vec_dot_iq3_xxs_q8_1( } #define VDR_IQ3_S_Q8_1_MMVQ 2 +#define VDR_IQ3_S_Q8_1_MMQ 2 // TODO: don't use lookup table for signs static __device__ __forceinline__ float vec_dot_iq3_s_q8_1( @@ -962,6 +939,9 @@ static __device__ __forceinline__ float vec_dot_iq3_s_q8_1( return d * sumi; } +#define VDR_IQ1_S_Q8_1_MMVQ 1 +#define VDR_IQ1_S_Q8_1_MMQ 1 + static __device__ __forceinline__ float vec_dot_iq1_s_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { const block_iq1_s * bq1 = (const block_iq1_s *) vbq + kbx; @@ -992,6 +972,9 @@ static __device__ __forceinline__ float vec_dot_iq1_s_q8_1( return d1q * (ds.x*sumi + ds.y*delta); } +#define VDR_IQ1_M_Q8_1_MMVQ 1 +#define VDR_IQ1_M_Q8_1_MMQ 1 + static __device__ __forceinline__ float vec_dot_iq1_m_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -1051,6 +1034,7 @@ static __device__ __forceinline__ int2 get_int_from_table_16(const int & q4) { } #define VDR_IQ4_NL_Q8_1_MMVQ 2 +#define VDR_IQ4_NL_Q8_1_MMQ 4 static __device__ __forceinline__ float vec_dot_iq4_nl_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { @@ -1074,6 +1058,7 @@ static __device__ __forceinline__ float vec_dot_iq4_nl_q8_1( } #define VDR_IQ4_XS_Q8_1_MMVQ 4 +#define VDR_IQ4_XS_Q8_1_MMQ 4 static __device__ __forceinline__ float vec_dot_iq4_xs_q8_1( const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int & kbx, const int & iqs) { diff --git a/ggml/src/ggml-sycl.cpp b/ggml/src/ggml-sycl.cpp index 76bad57e2..21006cd7b 100644 --- a/ggml/src/ggml-sycl.cpp +++ b/ggml/src/ggml-sycl.cpp @@ -49,7 +49,7 @@ bool ggml_backend_is_sycl(ggml_backend_t backend); int ggml_backend_sycl_get_device(ggml_backend_t backend); static bool ggml_backend_buffer_is_sycl_split(ggml_backend_buffer_t buffer); static inline int get_sycl_env(const char *env_name, int default_val); -static inline int get_work_group_size(const sycl::device& device); + void dev2dev_memcpy(sycl::queue &q_dst, sycl::queue &q_src, void *ptr_dst, const void *ptr_src, size_t size) { @@ -892,117 +892,6 @@ static void diag_mask_inf_f32(const float * x, float * dst, const int ncols, con dst[i] = x[i] - (col > n_past + row % rows_per_channel) * FLT_MAX; } - -template -static void soft_max_f32(const float * x, const float * mask, float * dst, const int ncols_par, - const int nrows_y, const float scale, const float max_bias, const float m0, - const float m1, uint32_t n_head_log2, const sycl::nd_item<3> &item_ct1, float *buf) { - const int ncols = ncols_template == 0 ? ncols_par : ncols_template; - - const int tid = item_ct1.get_local_id(2); - const int rowx = item_ct1.get_group(2); - const int rowy = rowx % nrows_y; // broadcast the mask (y) in the row dimension - - const int block_size = block_size_template == 0 ? item_ct1.get_local_range(2) : block_size_template; - - const int warp_id = item_ct1.get_local_id(2) / WARP_SIZE; - const int lane_id = item_ct1.get_local_id(2) % WARP_SIZE; - - float slope = 1.0f; - - // ALiBi - if (max_bias > 0.0f) { - const uint32_t h = rowx/nrows_y; // head index - - const float base = h < n_head_log2 ? m0 : m1; - const int exp = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1; - - slope = sycl::pow(base, float(exp)); - } - - float * vals = vals_smem ? buf + WARP_SIZE : dst + rowx*ncols; - float max_val = -INFINITY; - - for (int col0 = 0; col0 < ncols; col0 += block_size) { - const int col = col0 + tid; - - if (ncols_template == 0 && col >= ncols) { - break; - } - - const int ix = rowx*ncols + col; - const int iy = rowy*ncols + col; - - const float val = x[ix]*scale + (mask ? slope*mask[iy] : 0.0f); - - vals[col] = val; - max_val = sycl::max(max_val, val); - } - - // find the max value in the block - max_val = warp_reduce_max(max_val, item_ct1); - if (block_size > WARP_SIZE) { - if (warp_id == 0) { - buf[lane_id] = -INFINITY; - } - item_ct1.barrier(sycl::access::fence_space::local_space); - - if (lane_id == 0) { - buf[warp_id] = max_val; - } - item_ct1.barrier(sycl::access::fence_space::local_space); - - max_val = buf[lane_id]; - max_val = warp_reduce_max(max_val, item_ct1); - } - - float tmp = 0.f; - -#pragma unroll - for (int col0 = 0; col0 < ncols; col0 += block_size) { - const int col = col0 + tid; - if (ncols_template == 0 && col >= ncols) { - break; - } - - const float val = sycl::native::exp(vals[col] - max_val); - tmp += val; - vals[col] = val; - } - - // find the sum of exps in the block - tmp = warp_reduce_sum(tmp, item_ct1); - if (block_size > WARP_SIZE) { - item_ct1.barrier(sycl::access::fence_space::local_space); - if (warp_id == 0) { - buf[lane_id] = 0.f; - } - item_ct1.barrier(sycl::access::fence_space::local_space); - - if (lane_id == 0) { - buf[warp_id] = tmp; - } - item_ct1.barrier(sycl::access::fence_space::local_space); - - tmp = buf[lane_id]; - tmp = warp_reduce_sum(tmp, item_ct1); - } - - const float inv_sum = 1.f / tmp; - -#pragma unroll - for (int col0 = 0; col0 < ncols; col0 += block_size) { - const int col = col0 + tid; - - if (ncols_template == 0 && col >= ncols) { - return; - } - - const int idst = rowx*ncols + col; - dst[idst] = vals[col] * inv_sum; - } -} - static void scale_f32(const float * x, float * dst, const float scale, const int k, const sycl::nd_item<3> &item_ct1) { const int i = item_ct1.get_local_range(2) * item_ct1.get_group(2) + @@ -1890,106 +1779,6 @@ static void diag_mask_inf_f32_sycl(const float *x, float *dst, }); } -template -static void soft_max_f32_submitter(const float * x, const float * mask, float * dst, const int ncols_par, - const int nrows_y, const float scale, const float max_bias, const float m0, - const float m1, uint32_t n_head_log2, sycl::range<3> block_nums, sycl::range<3> block_dims, - const size_t n_local_scratch, queue_ptr stream) { - stream->submit([&](sycl::handler &cgh) { - sycl::local_accessor local_buf_acc(n_local_scratch, cgh); - - cgh.parallel_for( - sycl::nd_range<3>(block_nums * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { - soft_max_f32(x, mask, dst, ncols_par, - nrows_y, scale, max_bias, m0, - m1, n_head_log2, item_ct1, - local_buf_acc.get_pointer()); - }); - }); -} - -static void soft_max_f32_sycl(const float * x, const float * mask, - float * dst, const int ncols_x, const int nrows_x, - const int nrows_y, const float scale, const float max_bias, - queue_ptr stream) { - int nth = WARP_SIZE; - int max_block_size = get_work_group_size(stream->get_device()); - while (nth < ncols_x && nth < max_block_size) nth *= 2; - if (nth>max_block_size) nth = max_block_size; - - const sycl::range<3> block_dims(1, 1, nth); - const sycl::range<3> block_nums(1, 1, nrows_x); - const size_t n_local_scratch = (GGML_PAD(ncols_x, WARP_SIZE) + WARP_SIZE); - - const uint32_t n_head_kv = nrows_x/nrows_y; - const uint32_t n_head_log2 = 1u << (uint32_t) floorf(log2f((float) n_head_kv)); - - const float m0 = powf(2.0f, -(max_bias ) / n_head_log2); - const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2); - - const size_t local_mem_size = stream->get_device().get_info(); - if (n_local_scratch*sizeof(float) < local_mem_size) { - if (ncols_x > max_block_size) { - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - return; - } - switch (ncols_x) { - case 32: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 64: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 128: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 256: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 512: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 1024: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 2048: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - case 4096: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - default: - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, n_local_scratch, stream); - break; - } - } else { - soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, - max_bias, m0, m1, n_head_log2, block_nums, - block_dims, WARP_SIZE, stream); - } -} - template static void im2col_sycl(const float *x, T *dst, int IW, int IH, int OW, int OH, int KW, int KH, int IC, @@ -2156,6 +1945,8 @@ static ggml_sycl_device_info ggml_sycl_init() { info.devices[i].cc = 100 * prop.get_major_version() + 10 * prop.get_minor_version(); + + info.max_work_group_sizes[i] = prop.get_max_work_group_size(); } for (int id = 0; id < info.device_count; ++id) { @@ -3007,33 +2798,6 @@ inline void ggml_sycl_op_diag_mask_inf(ggml_backend_sycl_context & ctx, const gg (void) src1_dd; } -inline void ggml_sycl_op_soft_max(ggml_backend_sycl_context & ctx, const ggml_tensor *src0, - const ggml_tensor *src1, ggml_tensor *dst, - const float *src0_dd, const float *src1_dd, - float *dst_dd, - const queue_ptr &main_stream) { - - GGML_ASSERT(src0->type == GGML_TYPE_F32); - GGML_ASSERT( dst->type == GGML_TYPE_F32); - -#pragma message("TODO: add ggml_sycl_op_soft_max() F16 src1 support") -#pragma message("ref: https://github.com/ggerganov/llama.cpp/pull/5021") - GGML_ASSERT(!src1 || src1->type == GGML_TYPE_F32); // src1 contains mask and it is optional - - const int64_t ne00 = src0->ne[0]; - const int64_t nrows_x = ggml_nrows(src0); - const int64_t nrows_y = src0->ne[1]; - - float scale = 1.0f; - float max_bias = 0.0f; - - memcpy(&scale, dst->op_params + 0, sizeof(float)); - memcpy(&max_bias, dst->op_params + 1, sizeof(float)); - - soft_max_f32_sycl(src0_dd, src1 ? src1_dd : nullptr, dst_dd, ne00, - nrows_x, nrows_y, scale, max_bias, main_stream); -} - inline void ggml_sycl_op_scale(ggml_backend_sycl_context & ctx, const ggml_tensor *src0, const ggml_tensor *src1, ggml_tensor *dst, const float *src0_dd, const float *src1_dd, float *dst_dd, @@ -3729,10 +3493,6 @@ static void ggml_sycl_mul_mat_batched_sycl(ggml_backend_sycl_context & ctx, SYCL_CHECK(ggml_sycl_set_device(ctx.device)); queue_ptr main_stream = ctx.stream();; - bool no_mixed_dtypes = main_stream->get_backend() == sycl::backend::ext_oneapi_cuda || - main_stream->get_backend() == sycl::backend::ext_oneapi_hip; - - void * src0_ddq = src0->data; sycl::half *src0_as_f16 = (sycl::half *)src0_ddq; float * src1_ddf = (float *) src1->data; @@ -3750,15 +3510,10 @@ static void ggml_sycl_mul_mat_batched_sycl(ggml_backend_sycl_context & ctx, sycl::half *src1_f16 = src1->type == GGML_TYPE_F16 ? (sycl::half *)src1_ddf : src1_f16_alloc.get(); - ggml_sycl_pool_alloc dst_f16(ctx.pool()); char * dst_t; dpct::library_data_t cu_compute_type = dpct::library_data_t::real_float; dpct::library_data_t cu_data_type = dpct::library_data_t::real_float; - if (no_mixed_dtypes) { - cu_compute_type = dpct::library_data_t::real_half; - cu_data_type = dpct::library_data_t::real_half; - } // dst strides size_t nbd2 = dst->nb[2]; @@ -3767,26 +3522,10 @@ static void ggml_sycl_mul_mat_batched_sycl(ggml_backend_sycl_context & ctx, const float alpha_f32 = 1.0f; const float beta_f32 = 0.0f; - const sycl::half alpha_f16 = 1.0f; - const sycl::half beta_f16 = 0.0f; - const void * alpha = &alpha_f32; const void * beta = &beta_f32; - if (no_mixed_dtypes) { - alpha = &alpha_f16; - beta = &beta_f16; - } - - // TODO: Renable (dst->op_params[0] =! GGML_PREC_DEFAULT) pathway - // when oneMKL open source supports half, half, float, float: datatypes dst_t = (char *) dst_ddf; - if (no_mixed_dtypes) { - dst_t = (char *) dst_f16.alloc(ne_dst); - - nbd2 /= sizeof(float) / sizeof(sycl::half); - nbd3 /= sizeof(float) / sizeof(sycl::half); - } GGML_ASSERT(ne12 % ne02 == 0); GGML_ASSERT(ne13 % ne03 == 0); @@ -3848,11 +3587,6 @@ static void ggml_sycl_mul_mat_batched_sycl(ggml_backend_sycl_context & ctx, (void **)(ptrs_dst.get() + 0 * ne23), cu_data_type, ne01, ne23, cu_compute_type))); } - - if (no_mixed_dtypes) { - const to_fp32_sycl_t to_fp32_sycl = ggml_get_to_fp32_sycl(GGML_TYPE_F16); - to_fp32_sycl(dst_f16.get(), dst_ddf, ne_dst, main_stream); - } } catch (sycl::exception const &exc) { std::cerr << exc.what() << "Exception caught at file:" << __FILE__ @@ -5530,7 +5264,8 @@ GGML_CALL static bool ggml_backend_sycl_supports_op(ggml_backend_t backend, cons case GGML_OP_CONCAT: { ggml_type src0_type = op->src[0]->type; - return src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16; + int dim = op->op_params[0]; + return ggml_is_contiguous(op->src[0]) && ggml_is_contiguous(op->src[1]) && src0_type != GGML_TYPE_I32 && src0_type != GGML_TYPE_I16 && dim == 2; } break; case GGML_OP_DUP: case GGML_OP_NONE: diff --git a/ggml/src/ggml-sycl/backend.hpp b/ggml/src/ggml-sycl/backend.hpp index 3afa33919..2a789edfc 100644 --- a/ggml/src/ggml-sycl/backend.hpp +++ b/ggml/src/ggml-sycl/backend.hpp @@ -21,5 +21,6 @@ #include "mmvq.hpp" #include "rope.hpp" #include "norm.hpp" +#include "softmax.hpp" #endif // GGML_SYCL_BACKEND_HPP diff --git a/ggml/src/ggml-sycl/common.hpp b/ggml/src/ggml-sycl/common.hpp index 476d847ca..9a1c161b6 100644 --- a/ggml/src/ggml-sycl/common.hpp +++ b/ggml/src/ggml-sycl/common.hpp @@ -47,10 +47,6 @@ static int g_ggml_sycl_debug = 0; } \ }() -// #define DEBUG_SYCL_MALLOC - -static int g_work_group_size = 0; -// typedef sycl::half ggml_fp16_t; #define __SYCL_ARCH__ DPCT_COMPATIBILITY_TEMP #define VER_4VEC 610 // todo for hardward optimize. @@ -193,6 +189,8 @@ struct ggml_sycl_device_info { sycl_device_info devices[GGML_SYCL_MAX_DEVICES] = {}; std::array default_tensor_split = {}; + + int max_work_group_sizes[GGML_SYCL_MAX_DEVICES] = {0}; }; const ggml_sycl_device_info & ggml_sycl_info(); @@ -295,15 +293,6 @@ struct ggml_backend_sycl_context { } }; -// common host functions - -static inline int get_work_group_size(const sycl::device& device) { - dpct::device_info prop; - dpct::get_device_info(prop, device); - return prop.get_max_work_group_size(); -} - - // common device functions static __dpct_inline__ float warp_reduce_sum(float x, diff --git a/ggml/src/ggml-sycl/dmmv.cpp b/ggml/src/ggml-sycl/dmmv.cpp index 927819281..70a94fc16 100644 --- a/ggml/src/ggml-sycl/dmmv.cpp +++ b/ggml/src/ggml-sycl/dmmv.cpp @@ -3,6 +3,7 @@ #include "dequantize.hpp" #include "presets.hpp" + static void convert_f16(const void * vx, const int ib, const int iqs, dfloat2 & v){ const sycl::half *x = (const sycl::half *)vx; @@ -227,7 +228,7 @@ static void dequantize_mul_mat_vec_q2_k(const void *__restrict__ vx, // sum up partial sums and write back result #pragma unroll - for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { + for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) { tmp += dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); } @@ -346,7 +347,7 @@ static void dequantize_mul_mat_vec_q3_k(const void *__restrict__ vx, // sum up partial sums and write back result #pragma unroll - for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { + for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) { tmp += dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); } @@ -499,7 +500,7 @@ static void dequantize_mul_mat_vec_q4_k(const void *__restrict__ vx, // sum up partial sums and write back result #pragma unroll - for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { + for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) { tmp += dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); } @@ -633,7 +634,7 @@ static void dequantize_mul_mat_vec_q5_k(const void *__restrict__ vx, // sum up partial sums and write back result #pragma unroll - for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { + for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) { tmp += dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); } @@ -748,7 +749,7 @@ static void dequantize_mul_mat_vec_q6_k(const void * __restrict__ vx, const floa // sum up partial sums and write back result #pragma unroll - for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) { + for (int mask = QK_WARP_SIZE / 2; mask > 0; mask >>= 1) { tmp += dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask); } @@ -873,10 +874,10 @@ static void dequantize_mul_mat_vec_q2_K_sycl(const void *vx, const float *y, const int ny = 2; // very slightly faster than 1 even when K_QUANTS_PER_ITERATION = 2 const int block_num_y = (nrows + ny - 1) / ny; const sycl::range<3> block_nums(1, 1, block_num_y); - const sycl::range<3> block_dims(1, ny, WARP_SIZE); + const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE); stream->parallel_for( sycl::nd_range<3>(block_nums * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] { dequantize_mul_mat_vec_q2_k(vx, y, dst, ncols, nrows, item_ct1); }); } @@ -889,10 +890,10 @@ static void dequantize_mul_mat_vec_q3_K_sycl(const void *vx, const float *y, const int ny = 2 / K_QUANTS_PER_ITERATION; const int block_num_y = (nrows + ny - 1) / ny; const sycl::range<3> block_nums(1, 1, block_num_y); - const sycl::range<3> block_dims(1, ny, WARP_SIZE); + const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE); stream->parallel_for( sycl::nd_range<3>(block_nums * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] { dequantize_mul_mat_vec_q3_k(vx, y, dst, ncols, nrows, item_ct1); }); } @@ -905,10 +906,10 @@ static void dequantize_mul_mat_vec_q4_K_sycl(const void *vx, const float *y, const int ny = 2 / K_QUANTS_PER_ITERATION; const int block_num_y = (nrows + ny - 1) / ny; const sycl::range<3> block_nums(1, 1, block_num_y); - const sycl::range<3> block_dims(1, ny, WARP_SIZE); + const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE); stream->parallel_for( sycl::nd_range<3>(block_nums * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] { dequantize_mul_mat_vec_q4_k(vx, y, dst, ncols, nrows, item_ct1); }); } @@ -918,10 +919,10 @@ static void dequantize_mul_mat_vec_q5_K_sycl(const void *vx, const float *y, const int nrows, dpct::queue_ptr stream) { GGML_ASSERT(ncols % QK_K == 0); - const sycl::range<3> block_dims(1, 1, WARP_SIZE); + const sycl::range<3> block_dims(1, 1, QK_WARP_SIZE); stream->parallel_for( sycl::nd_range<3>(sycl::range<3>(1, 1, nrows) * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] { dequantize_mul_mat_vec_q5_k(vx, y, dst, ncols, item_ct1); }); } @@ -934,10 +935,10 @@ static void dequantize_mul_mat_vec_q6_K_sycl(const void *vx, const float *y, const int ny = 2 / K_QUANTS_PER_ITERATION; const int block_num_y = (nrows + ny - 1) / ny; const sycl::range<3> block_nums(1, 1, block_num_y); - const sycl::range<3> block_dims(1, ny, WARP_SIZE); + const sycl::range<3> block_dims(1, ny, QK_WARP_SIZE); stream->parallel_for( sycl::nd_range<3>(block_nums * block_dims, block_dims), - [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(QK_WARP_SIZE)]] { dequantize_mul_mat_vec_q6_k(vx, y, dst, ncols, nrows, item_ct1); }); } diff --git a/ggml/src/ggml-sycl/dpct/helper.hpp b/ggml/src/ggml-sycl/dpct/helper.hpp index 1ff297218..31df1cb9e 100644 --- a/ggml/src/ggml-sycl/dpct/helper.hpp +++ b/ggml/src/ggml-sycl/dpct/helper.hpp @@ -255,7 +255,7 @@ namespace dpct void set_pitch(size_t pitch) { _pitch = pitch; } size_t get_x() { return _x; } - void set_x(size_t x) { _x = x; }; + void set_x(size_t x) { _x = x; } size_t get_y() { return _y; } void set_y(size_t y) { _y = y; } @@ -1056,7 +1056,7 @@ namespace dpct #error "Only support Windows and Linux." #endif next_free = mapped_address_space; - }; + } public: using buffer_id_t = int; @@ -1077,7 +1077,7 @@ namespace dpct #else #error "Only support Windows and Linux." #endif - }; + } mem_mgr(const mem_mgr &) = delete; mem_mgr &operator=(const mem_mgr &) = delete; @@ -2426,6 +2426,7 @@ namespace dpct b, ldb, beta, c, ldc, batch_size); break; } +#endif case detail::get_type_combination_id( library_data_t::real_int8, library_data_t::real_int8, library_data_t::real_int32, library_data_t::real_int32): @@ -2458,7 +2459,6 @@ namespace dpct batch_size); break; } -#endif case detail::get_type_combination_id( library_data_t::real_half, library_data_t::real_half, library_data_t::real_half, library_data_t::real_float): @@ -2595,6 +2595,7 @@ namespace dpct stride_c, batch_size); break; } +#endif case detail::get_type_combination_id( library_data_t::real_int8, library_data_t::real_int8, library_data_t::real_int32, library_data_t::real_int32): @@ -2623,7 +2624,6 @@ namespace dpct beta, c, ldc, stride_c, batch_size); break; } -#endif case detail::get_type_combination_id( library_data_t::real_half, library_data_t::real_half, library_data_t::real_half, library_data_t::real_float): diff --git a/ggml/src/ggml-sycl/norm.cpp b/ggml/src/ggml-sycl/norm.cpp index a77f7852c..e0c5dfeca 100644 --- a/ggml/src/ggml-sycl/norm.cpp +++ b/ggml/src/ggml-sycl/norm.cpp @@ -57,6 +57,7 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con const int nwarps = nthreads / WARP_SIZE; assert(nwarps % WARP_SIZE == 0); start += item_ct1.get_local_id(2); + int nreduce = nwarps / WARP_SIZE; if (end >= ne_elements) { end = ne_elements; @@ -87,7 +88,6 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con */ item_ct1.barrier(); tmp = 0.f; - int nreduce = nwarps / WARP_SIZE; for (size_t i = 0; i < nreduce; i += 1) { tmp += s_sum[lane_id + i * WARP_SIZE]; @@ -122,7 +122,11 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con better performance if there is no access to global memory. */ item_ct1.barrier(); - tmp = s_sum[lane_id]; + tmp = 0.f; + for (size_t i = 0; i < nreduce; i += 1) + { + tmp += s_sum[lane_id + i * WARP_SIZE]; + } tmp = warp_reduce_sum(tmp, item_ct1); } @@ -181,7 +185,7 @@ static void rms_norm_f32(const float* x, float* dst, const int ncols, const floa static void norm_f32_sycl(const float* x, float* dst, const int ncols, const int nrows, const float eps, - queue_ptr stream) { + queue_ptr stream, int device) { GGML_ASSERT(ncols % WARP_SIZE == 0); if (ncols < 1024) { const sycl::range<3> block_dims(1, 1, WARP_SIZE); @@ -197,7 +201,7 @@ static void norm_f32_sycl(const float* x, float* dst, const int ncols, }); } else { - const int work_group_size = get_work_group_size(stream->get_device()); + const int work_group_size = ggml_sycl_info().max_work_group_sizes[device]; const sycl::range<3> block_dims(1, 1, work_group_size); /* DPCT1049:17: The work-group size passed to the SYCL kernel may exceed @@ -222,7 +226,7 @@ static void norm_f32_sycl(const float* x, float* dst, const int ncols, static void group_norm_f32_sycl(const float* x, float* dst, const int num_groups, const int group_size, - const int ne_elements, queue_ptr stream) { + const int ne_elements, queue_ptr stream, int device) { static const float eps = 1e-6f; if (group_size < 1024) { const sycl::range<3> block_dims(1, 1, WARP_SIZE); @@ -240,7 +244,7 @@ static void group_norm_f32_sycl(const float* x, float* dst, }); } else { - const int work_group_size = get_work_group_size(stream->get_device()); + const int work_group_size = ggml_sycl_info().max_work_group_sizes[device]; const sycl::range<3> block_dims(1, 1, work_group_size); /* DPCT1049:18: The work-group size passed to the SYCL kernel may exceed @@ -269,7 +273,7 @@ static void group_norm_f32_sycl(const float* x, float* dst, static void rms_norm_f32_sycl(const float* x, float* dst, const int ncols, const int nrows, const float eps, - queue_ptr stream) { + queue_ptr stream, int device) { GGML_ASSERT(ncols % WARP_SIZE == 0); // printf("%s ncols=%d, nrows=%d, WARP_SIZE=%d\n", __func__, ncols, nrows, WARP_SIZE); if (ncols < 1024) { @@ -286,7 +290,7 @@ static void rms_norm_f32_sycl(const float* x, float* dst, const int ncols, }); } else { - const int work_group_size = get_work_group_size(stream->get_device()); + const int work_group_size = ggml_sycl_info().max_work_group_sizes[device]; const sycl::range<3> block_dims(1, 1, work_group_size); /* DPCT1049:19: The work-group size passed to the SYCL kernel may exceed @@ -322,7 +326,7 @@ void ggml_sycl_op_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* src0, float eps; memcpy(&eps, dst->op_params, sizeof(float)); - norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream); + norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream, ctx.device); (void)src1; (void)dst; @@ -340,7 +344,7 @@ void ggml_sycl_op_group_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* int num_groups = dst->op_params[0]; int group_size = src0->ne[0] * src0->ne[1] * ((src0->ne[2] + num_groups - 1) / num_groups); - group_norm_f32_sycl(src0_dd, dst_dd, num_groups, group_size, src0->ne[0] * src0->ne[1] * src0->ne[2], main_stream); + group_norm_f32_sycl(src0_dd, dst_dd, num_groups, group_size, src0->ne[0] * src0->ne[1] * src0->ne[2], main_stream, ctx.device); (void)src1; (void)dst; @@ -362,7 +366,7 @@ void ggml_sycl_op_rms_norm(ggml_backend_sycl_context& ctx, const ggml_tensor* sr float eps; memcpy(&eps, dst->op_params, sizeof(float)); - rms_norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream); + rms_norm_f32_sycl(src0_dd, dst_dd, ne00, nrows, eps, main_stream, ctx.device); (void)src1; (void)dst; diff --git a/ggml/src/ggml-sycl/presets.hpp b/ggml/src/ggml-sycl/presets.hpp index c09c75dc7..15ddcac1f 100644 --- a/ggml/src/ggml-sycl/presets.hpp +++ b/ggml/src/ggml-sycl/presets.hpp @@ -62,4 +62,5 @@ static_assert(K_QUANTS_PER_ITERATION == 1 || K_QUANTS_PER_ITERATION == 2, "K_QUA #define MUL_MAT_SRC1_COL_STRIDE 128 +#define QK_WARP_SIZE 32 #endif // GGML_SYCL_PRESETS_HPP diff --git a/ggml/src/ggml-sycl/softmax.cpp b/ggml/src/ggml-sycl/softmax.cpp new file mode 100644 index 000000000..e624b6ba3 --- /dev/null +++ b/ggml/src/ggml-sycl/softmax.cpp @@ -0,0 +1,250 @@ +#include "norm.hpp" + +template +static void soft_max_f32(const float * x, const float * mask, float * dst, const int ncols_par, + const int nrows_y, const float scale, const float max_bias, const float m0, + const float m1, uint32_t n_head_log2, const sycl::nd_item<3> &item_ct1, float *buf) { + const int ncols = ncols_template == 0 ? ncols_par : ncols_template; + + const int tid = item_ct1.get_local_id(2); + const int rowx = item_ct1.get_group(2); + const int rowy = rowx % nrows_y; // broadcast the mask (y) in the row dimension + + const int block_size = block_size_template == 0 ? item_ct1.get_local_range(2) : block_size_template; + + const int warp_id = item_ct1.get_local_id(2) / WARP_SIZE; + const int lane_id = item_ct1.get_local_id(2) % WARP_SIZE; + const int nthreads = block_size; + const int nwarps = nthreads / WARP_SIZE; + int nreduce = nwarps / WARP_SIZE; + float slope = 1.0f; + + // ALiBi + if (max_bias > 0.0f) { + const uint32_t h = rowx/nrows_y; // head index + + const float base = h < n_head_log2 ? m0 : m1; + const int exp = h < n_head_log2 ? h + 1 : 2*(h - n_head_log2) + 1; + + slope = sycl::pow(base, float(exp)); + } + + float *vals = vals_smem ? buf + std::max(nwarps, WARP_SIZE) : dst + rowx * ncols; + float max_val = -INFINITY; + + for (int col0 = 0; col0 < ncols; col0 += block_size) { + const int col = col0 + tid; + + if (ncols_template == 0 && col >= ncols) { + break; + } + + const int ix = rowx*ncols + col; + const int iy = rowy*ncols + col; + + const float val = x[ix]*scale + (mask ? slope*mask[iy] : 0.0f); + + vals[col] = val; + max_val = sycl::max(max_val, val); + } + + // find the max value in the block + max_val = warp_reduce_max(max_val, item_ct1); + if (block_size > WARP_SIZE) { + if (warp_id == 0) { + buf[lane_id] = -INFINITY; + for (size_t i = 1; i < nreduce; i += 1) + buf[lane_id + i * WARP_SIZE] = -INFINITY; + } + item_ct1.barrier(sycl::access::fence_space::local_space); + + if (lane_id == 0) { + buf[warp_id] = max_val; + } + item_ct1.barrier(sycl::access::fence_space::local_space); + max_val = buf[lane_id]; + for (size_t i = 1; i < nreduce; i += 1) + { + max_val = std::max(max_val, buf[lane_id + i * WARP_SIZE]); + } + max_val = warp_reduce_max(max_val, item_ct1); + } + + float tmp = 0.f; +#pragma unroll + for (int col0 = 0; col0 < ncols; col0 += block_size) { + const int col = col0 + tid; + if (ncols_template == 0 && col >= ncols) { + break; + } + + const float val = sycl::native::exp(vals[col] - max_val); + tmp += val; + vals[col] = val; + } + + // find the sum of exps in the block + tmp = warp_reduce_sum(tmp, item_ct1); + if (block_size > WARP_SIZE) { + item_ct1.barrier(sycl::access::fence_space::local_space); + if (warp_id == 0) { + buf[lane_id] = 0.f; + for (size_t i = 1; i < nreduce; i += 1) + buf[lane_id + i * WARP_SIZE] = 0.f; + } + item_ct1.barrier(sycl::access::fence_space::local_space); + + if (lane_id == 0) { + buf[warp_id] = tmp; + } + item_ct1.barrier(sycl::access::fence_space::local_space); + + tmp = buf[lane_id]; + for (size_t i = 1; i < nreduce; i += 1) + { + tmp += buf[lane_id + i * WARP_SIZE]; + } + tmp = warp_reduce_sum(tmp, item_ct1); + } + + const float inv_sum = 1.f / tmp; + +#pragma unroll + for (int col0 = 0; col0 < ncols; col0 += block_size) { + const int col = col0 + tid; + + if (ncols_template == 0 && col >= ncols) { + return; + } + + const int idst = rowx*ncols + col; + dst[idst] = vals[col] * inv_sum; + } +} + +template +static void soft_max_f32_submitter(const float * x, const float * mask, float * dst, const int ncols_par, + const int nrows_y, const float scale, const float max_bias, const float m0, + const float m1, uint32_t n_head_log2, sycl::range<3> block_nums, sycl::range<3> block_dims, + const size_t n_local_scratch, queue_ptr stream) { + stream->submit([&](sycl::handler &cgh) { + sycl::local_accessor local_buf_acc(n_local_scratch, cgh); + + cgh.parallel_for( + sycl::nd_range<3>(block_nums * block_dims, block_dims), + [=](sycl::nd_item<3> item_ct1) [[intel::reqd_sub_group_size(WARP_SIZE)]] { + soft_max_f32(x, mask, dst, ncols_par, + nrows_y, scale, max_bias, m0, + m1, n_head_log2, item_ct1, + local_buf_acc.get_pointer()); + }); + }); +} + +static void soft_max_f32_sycl(const float * x, const float * mask, + float * dst, const int ncols_x, const int nrows_x, + const int nrows_y, const float scale, const float max_bias, + queue_ptr stream, int device) { + int nth = WARP_SIZE; + int max_block_size = ggml_sycl_info().max_work_group_sizes[device]; + while (nth < ncols_x && nth < max_block_size) nth *= 2; + if (nth>max_block_size) nth = max_block_size; + + const sycl::range<3> block_dims(1, 1, nth); + const sycl::range<3> block_nums(1, 1, nrows_x); + const size_t n_local_scratch = (GGML_PAD(ncols_x, WARP_SIZE) + WARP_SIZE); + + const uint32_t n_head_kv = nrows_x/nrows_y; + const uint32_t n_head_log2 = 1u << (uint32_t) floorf(log2f((float) n_head_kv)); + + const float m0 = powf(2.0f, -(max_bias ) / n_head_log2); + const float m1 = powf(2.0f, -(max_bias / 2.0f) / n_head_log2); + + const size_t local_mem_size = stream->get_device().get_info(); + if (n_local_scratch*sizeof(float) < local_mem_size) { + if (ncols_x > max_block_size) { + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + return; + } + switch (ncols_x) { + case 32: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 64: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 128: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 256: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 512: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 1024: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 2048: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + case 4096: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + default: + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, n_local_scratch, stream); + break; + } + } else { + soft_max_f32_submitter(x, mask, dst, ncols_x, nrows_y, scale, + max_bias, m0, m1, n_head_log2, block_nums, + block_dims, WARP_SIZE, stream); + } +} + +void ggml_sycl_op_soft_max(ggml_backend_sycl_context & ctx, const ggml_tensor *src0, + const ggml_tensor *src1, ggml_tensor *dst, + const float *src0_dd, const float *src1_dd, + float *dst_dd, + const queue_ptr &main_stream) { + + GGML_ASSERT(src0->type == GGML_TYPE_F32); + GGML_ASSERT( dst->type == GGML_TYPE_F32); + +#pragma message("TODO: add ggml_sycl_op_soft_max() F16 src1 support") +#pragma message("ref: https://github.com/ggerganov/llama.cpp/pull/5021") + GGML_ASSERT(!src1 || src1->type == GGML_TYPE_F32); // src1 contains mask and it is optional + + const int64_t ne00 = src0->ne[0]; + const int64_t nrows_x = ggml_nrows(src0); + const int64_t nrows_y = src0->ne[1]; + + float scale = 1.0f; + float max_bias = 0.0f; + + memcpy(&scale, dst->op_params + 0, sizeof(float)); + memcpy(&max_bias, dst->op_params + 1, sizeof(float)); + + soft_max_f32_sycl(src0_dd, src1 ? src1_dd : nullptr, dst_dd, ne00, + nrows_x, nrows_y, scale, max_bias, main_stream, ctx.device); +} diff --git a/ggml/src/ggml-sycl/softmax.hpp b/ggml/src/ggml-sycl/softmax.hpp new file mode 100644 index 000000000..bdb8f712e --- /dev/null +++ b/ggml/src/ggml-sycl/softmax.hpp @@ -0,0 +1,24 @@ +// +// MIT license +// Copyright (C) 2024 Intel Corporation +// SPDX-License-Identifier: MIT +// + +// +// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. +// See https://llvm.org/LICENSE.txt for license information. +// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception +// + +#ifndef GGML_SYCL_SOFTMAX_HPP +#define GGML_SYCL_SOFTMAX_HPP + +#include "common.hpp" + +void ggml_sycl_op_soft_max(ggml_backend_sycl_context &ctx, const ggml_tensor *src0, + const ggml_tensor *src1, ggml_tensor *dst, + const float *src0_dd, const float *src1_dd, + float *dst_dd, + const queue_ptr &main_stream); + +#endif // GGML_SYCL_SOFTMAX_HPP diff --git a/ggml/src/ggml.c b/ggml/src/ggml.c index f5502afbe..bc91ac3a7 100644 --- a/ggml/src/ggml.c +++ b/ggml/src/ggml.c @@ -5312,7 +5312,7 @@ void ggml_mul_mat_set_prec( as -> [cols, rows, n_expert] ids -> [n_experts_used, n_tokens] (i32) b -> [cols, n_expert_used, n_tokens] - c -> [cols, n_expert_used, n_tokens] + c -> [rows, n_expert_used, n_tokens] in b, n_experts_used can be broadcasted to match the n_expert_used of ids diff --git a/gguf-py/README.md b/gguf-py/README.md index a04c22759..bc46d6e1d 100644 --- a/gguf-py/README.md +++ b/gguf-py/README.md @@ -3,7 +3,7 @@ This is a Python package for writing binary files in the [GGUF](https://github.com/ggerganov/ggml/pull/302) (GGML Universal File) format. -See [convert-llama-hf-to-gguf.py](https://github.com/ggerganov/llama.cpp/blob/master/convert-hf-to-gguf.py) +See [convert_hf_to_gguf.py](https://github.com/ggerganov/llama.cpp/blob/master/convert_hf_to_gguf.py) as an example for its usage. ## Installation @@ -15,13 +15,13 @@ pip install gguf [examples/writer.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/examples/writer.py) — Generates `example.gguf` in the current directory to demonstrate generating a GGUF file. Note that this file cannot be used as a model. -[scripts/gguf-dump.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf-dump.py) — Dumps a GGUF file's metadata to the console. +[scripts/gguf_dump.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf_dump.py) — Dumps a GGUF file's metadata to the console. -[scripts/gguf-set-metadata.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf-set-metadata.py) — Allows changing simple metadata values in a GGUF file by key. +[scripts/gguf_set_metadata.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf_set_metadata.py) — Allows changing simple metadata values in a GGUF file by key. -[scripts/gguf-convert-endian.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf-convert-endian.py) — Allows converting the endianness of GGUF files. +[scripts/gguf_convert_endian.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf_convert_endian.py) — Allows converting the endianness of GGUF files. -[scripts/gguf-new-metadata.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf-new-metadata.py) — Copies a GGUF file with added/modified/removed metadata values. +[scripts/gguf_new_metadata.py](https://github.com/ggerganov/llama.cpp/blob/master/gguf-py/scripts/gguf_new_metadata.py) — Copies a GGUF file with added/modified/removed metadata values. ## Development Maintainers who participate in development of this package are advised to install it in editable mode: diff --git a/gguf-py/gguf/constants.py b/gguf-py/gguf/constants.py index 8c6ef61e9..a95a44237 100644 --- a/gguf-py/gguf/constants.py +++ b/gguf-py/gguf/constants.py @@ -159,6 +159,7 @@ class MODEL_ARCH(IntEnum): COMMAND_R = auto() DBRX = auto() OLMO = auto() + OPENELM = auto() ARCTIC = auto() DEEPSEEK2 = auto() CHATGLM = auto() @@ -285,6 +286,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = { MODEL_ARCH.COMMAND_R: "command-r", MODEL_ARCH.DBRX: "dbrx", MODEL_ARCH.OLMO: "olmo", + MODEL_ARCH.OPENELM: "openelm", MODEL_ARCH.ARCTIC: "arctic", MODEL_ARCH.DEEPSEEK2: "deepseek2", MODEL_ARCH.CHATGLM: "chatglm", @@ -862,6 +864,19 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = { MODEL_TENSOR.FFN_DOWN, MODEL_TENSOR.FFN_UP, ], + MODEL_ARCH.OPENELM: [ + MODEL_TENSOR.TOKEN_EMBD, + MODEL_TENSOR.OUTPUT_NORM, + MODEL_TENSOR.ATTN_NORM, + MODEL_TENSOR.ATTN_QKV, + MODEL_TENSOR.ATTN_Q_NORM, + MODEL_TENSOR.ATTN_K_NORM, + MODEL_TENSOR.ATTN_OUT, + MODEL_TENSOR.FFN_NORM, + MODEL_TENSOR.FFN_GATE, + MODEL_TENSOR.FFN_DOWN, + MODEL_TENSOR.FFN_UP, + ], MODEL_ARCH.ARCTIC: [ MODEL_TENSOR.TOKEN_EMBD, MODEL_TENSOR.OUTPUT_NORM, diff --git a/gguf-py/gguf/gguf_writer.py b/gguf-py/gguf/gguf_writer.py index 75a8b2636..cf9554162 100644 --- a/gguf-py/gguf/gguf_writer.py +++ b/gguf-py/gguf/gguf_writer.py @@ -480,8 +480,11 @@ class GGUFWriter: def add_leading_dense_block_count(self, length: int) -> None: self.add_uint32(Keys.LLM.LEADING_DENSE_BLOCK_COUNT.format(arch=self.arch), length) - def add_feed_forward_length(self, length: int) -> None: - self.add_uint32(Keys.LLM.FEED_FORWARD_LENGTH.format(arch=self.arch), length) + def add_feed_forward_length(self, length: int | Sequence[int]) -> None: + if isinstance(length, int): + self.add_uint32(Keys.LLM.FEED_FORWARD_LENGTH.format(arch=self.arch), length) + else: + self.add_array(Keys.LLM.FEED_FORWARD_LENGTH.format(arch=self.arch), length) def add_expert_feed_forward_length(self, length: int) -> None: self.add_uint32(Keys.LLM.EXPERT_FEED_FORWARD_LENGTH.format(arch=self.arch), length) @@ -495,11 +498,17 @@ class GGUFWriter: def add_decoder_start_token_id(self, id: int) -> None: self.add_uint32(Keys.LLM.DECODER_START_TOKEN_ID.format(arch=self.arch), id) - def add_head_count(self, count: int) -> None: - self.add_uint32(Keys.Attention.HEAD_COUNT.format(arch=self.arch), count) + def add_head_count(self, count: int | Sequence[int]) -> None: + if isinstance(count, int): + self.add_uint32(Keys.Attention.HEAD_COUNT.format(arch=self.arch), count) + else: + self.add_array(Keys.Attention.HEAD_COUNT.format(arch=self.arch), count) - def add_head_count_kv(self, count: int) -> None: - self.add_uint32(Keys.Attention.HEAD_COUNT_KV.format(arch=self.arch), count) + def add_head_count_kv(self, count: int | Sequence[int]) -> None: + if isinstance(count, int): + self.add_uint32(Keys.Attention.HEAD_COUNT_KV.format(arch=self.arch), count) + else: + self.add_array(Keys.Attention.HEAD_COUNT_KV.format(arch=self.arch), count) def add_key_length(self, length: int) -> None: self.add_uint32(Keys.Attention.KEY_LENGTH.format(arch=self.arch), length) diff --git a/gguf-py/gguf/tensor_mapping.py b/gguf-py/gguf/tensor_mapping.py index af311fb4e..7264240f5 100644 --- a/gguf-py/gguf/tensor_mapping.py +++ b/gguf-py/gguf/tensor_mapping.py @@ -25,6 +25,7 @@ class TensorNameMap: "backbone.embeddings", # mamba-hf "transformer.in_out_embed", # Grok "embedding.word_embeddings", # chatglm + "transformer.token_embeddings", # openelm "shared", # t5 ), @@ -38,6 +39,7 @@ class TensorNameMap: "word_embeddings_layernorm", # bloom "embeddings.LayerNorm", # bert "emb_ln", # nomic-bert + "transformer.norm", # openelm ), # Position embeddings @@ -72,6 +74,7 @@ class TensorNameMap: "backbone.norm_f", # mamba "transformer.rms_norm", # Grok "encoder.final_layernorm", # chatglm + "transformer.norm", # openelm ), # Rope frequencies @@ -103,6 +106,7 @@ class TensorNameMap: "transformer.decoder_layer.{bid}.rms_norm", # Grok "transformer.blocks.{bid}.norm_attn_norm.norm_1", # dbrx "encoder.layers.{bid}.input_layernorm", # chatglm + "transformer.layers.{bid}.attn_norm", # openelm ), # Attention norm 2 @@ -126,6 +130,7 @@ class TensorNameMap: "encoder.layers.{bid}.attn.Wqkv", # nomic-bert "model.layers.{bid}.self_attn.qkv_proj", # phi3 "encoder.layers.{bid}.self_attention.query_key_value", # chatglm + "transformer.layers.{bid}.attn.qkv_proj", # openelm ), # Attention query @@ -184,6 +189,7 @@ class TensorNameMap: "transformer.decoder_layer.{bid}.multi_head_attention.linear", # Grok "transformer.blocks.{bid}.norm_attn_norm.attn.out_proj", # dbrx "encoder.layers.{bid}.self_attention.dense", # chatglm + "transformer.layers.{bid}.attn.out_proj", # openelm ), # Attention output norm @@ -220,6 +226,7 @@ class TensorNameMap: "model.layers.{bid}.ffn_norm", # internlm2 "transformer.decoder_layer.{bid}.rms_norm_2", # Grok "encoder.layers.{bid}.post_attention_layernorm", # chatglm + "transformer.layers.{bid}.ffn_norm", # openelm ), # Post feed-forward norm @@ -336,6 +343,7 @@ class TensorNameMap: "encoder.layers.{bid}.mlp.fc2", # nomic-bert "model.layers.{bid}.mlp.c_proj", # starcoder2 "encoder.layer.{bid}.mlp.wo", # jina-bert-v2 + "transformer.layers.{bid}.ffn.proj_2", # openelm "model.layers.{bid}.residual_mlp.w2", # arctic "encoder.layer.{bid}.mlp.down_layer", # jina-bert-v2 "encoder.layers.{bid}.mlp.dense_4h_to_h", # chatglm @@ -358,7 +366,8 @@ class TensorNameMap: "model.layers.{bid}.self_attn.q_layernorm", # persimmon "model.layers.{bid}.self_attn.q_norm", # cohere "transformer.blocks.{bid}.attn.q_ln", # sea-lion - "encoder.layer.{bid}.attention.self.layer_norm_q" # jina-bert-v2 + "encoder.layer.{bid}.attention.self.layer_norm_q", # jina-bert-v2 + "transformer.layers.{bid}.attn.q_norm", # openelm ), MODEL_TENSOR.ATTN_K_NORM: ( @@ -366,7 +375,8 @@ class TensorNameMap: "model.layers.{bid}.self_attn.k_layernorm", # persimmon "model.layers.{bid}.self_attn.k_norm", # cohere "transformer.blocks.{bid}.attn.k_ln", # sea-lion - "encoder.layer.{bid}.attention.self.layer_norm_k" # jina-bert-v2 + "encoder.layer.{bid}.attention.self.layer_norm_k", # jina-bert-v2 + "transformer.layers.{bid}.attn.k_norm", # openelm ), MODEL_TENSOR.ROPE_FREQS: ( diff --git a/gguf-py/scripts/__init__.py b/gguf-py/scripts/__init__.py index 1ad45639a..f9d29cb69 100644 --- a/gguf-py/scripts/__init__.py +++ b/gguf-py/scripts/__init__.py @@ -1,13 +1,4 @@ -import os - -from importlib import import_module - - -os.environ["NO_LOCAL_GGUF"] = "TRUE" - -gguf_convert_endian_entrypoint = import_module("scripts.gguf-convert-endian").main -gguf_dump_entrypoint = import_module("scripts.gguf-dump").main -gguf_set_metadata_entrypoint = import_module("scripts.gguf-set-metadata").main -gguf_new_metadata_entrypoint = import_module("scripts.gguf-new-metadata").main - -del import_module, os +from .gguf_convert_endian import main as gguf_convert_endian_entrypoint +from .gguf_dump import main as gguf_dump_entrypoint +from .gguf_set_metadata import main as gguf_set_metadata_entrypoint +from .gguf_new_metadata import main as gguf_new_metadata_entrypoint diff --git a/gguf-py/scripts/gguf-convert-endian.py b/gguf-py/scripts/gguf_convert_endian.py similarity index 100% rename from gguf-py/scripts/gguf-convert-endian.py rename to gguf-py/scripts/gguf_convert_endian.py diff --git a/gguf-py/scripts/gguf-dump.py b/gguf-py/scripts/gguf_dump.py similarity index 100% rename from gguf-py/scripts/gguf-dump.py rename to gguf-py/scripts/gguf_dump.py diff --git a/gguf-py/scripts/gguf-new-metadata.py b/gguf-py/scripts/gguf_new_metadata.py similarity index 100% rename from gguf-py/scripts/gguf-new-metadata.py rename to gguf-py/scripts/gguf_new_metadata.py diff --git a/gguf-py/scripts/gguf-set-metadata.py b/gguf-py/scripts/gguf_set_metadata.py similarity index 100% rename from gguf-py/scripts/gguf-set-metadata.py rename to gguf-py/scripts/gguf_set_metadata.py diff --git a/include/llama.h b/include/llama.h index 16095d9a7..bb4b05ba6 100644 --- a/include/llama.h +++ b/include/llama.h @@ -182,6 +182,12 @@ extern "C" { LLAMA_POOLING_TYPE_LAST = 3, }; + enum llama_attention_type { + LLAMA_ATTENTION_TYPE_UNSPECIFIED = -1, + LLAMA_ATTENTION_TYPE_CAUSAL = 0, + LLAMA_ATTENTION_TYPE_NON_CAUSAL = 1, + }; + enum llama_split_mode { LLAMA_SPLIT_MODE_NONE = 0, // single GPU LLAMA_SPLIT_MODE_LAYER = 1, // split layers and KV across GPUs @@ -299,6 +305,7 @@ extern "C" { enum llama_rope_scaling_type rope_scaling_type; // RoPE scaling type, from `enum llama_rope_scaling_type` enum llama_pooling_type pooling_type; // whether to pool (sum) embedding results by sequence id + enum llama_attention_type attention_type; // attention type to use for embeddings // ref: https://github.com/ggerganov/llama.cpp/pull/2054 float rope_freq_base; // RoPE base frequency, 0 = from model @@ -487,6 +494,13 @@ extern "C" { // Get a llama model tensor LLAMA_API struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const char * name); + // Returns true if the model contains an encoder that requires llama_encode() call + LLAMA_API bool llama_model_has_encoder(const struct llama_model * model); + + // For encoder-decoder models, this function returns id of the token that must be provided + // to the decoder to start generating output sequence. For other models, it returns -1. + LLAMA_API llama_token llama_model_decoder_start_token(const struct llama_model * model); + // Returns 0 on success LLAMA_API uint32_t llama_model_quantize( const char * fname_inp, @@ -772,6 +786,14 @@ extern "C" { // Frees a batch of tokens allocated with llama_batch_init() LLAMA_API void llama_batch_free(struct llama_batch batch); + // Processes a batch of tokens with the ecoder part of the encoder-decoder model. + // Stores the encoder output internally for later use by the decoder cross-attention layers. + // 0 - success + // < 0 - error + LLAMA_API int32_t llama_encode( + struct llama_context * ctx, + struct llama_batch batch); + // Positive return values does not mean a fatal error, but rather a warning. // 0 - success // 1 - could not find a KV slot for the batch (try reducing the size of the batch or increase the context) @@ -884,6 +906,7 @@ extern "C" { /// @param tokens The tokens pointer must be large enough to hold the resulting tokens. /// @return Returns the number of tokens on success, no more than n_tokens_max /// @return Returns a negative number on failure - the number of tokens that would have been returned + /// @param add_special Allow to add BOS and EOS tokens if model is configured to do so. /// @param parse_special Allow tokenizing special and/or control tokens which otherwise are not exposed and treated /// as plaintext. Does not insert a leading space. LLAMA_API int32_t llama_tokenize( @@ -898,15 +921,31 @@ extern "C" { // Token Id -> Piece. // Uses the vocabulary in the provided context. // Does not write null terminator to the buffer. - // User code is responsible to remove the leading whitespace of the first non-BOS token when decoding multiple tokens. + // User can skip up to 'lstrip' leading spaces before copying (useful when encoding/decoding multiple tokens with 'add_space_prefix') // @param special If true, special tokens are rendered in the output. LLAMA_API int32_t llama_token_to_piece( const struct llama_model * model, llama_token token, char * buf, int32_t length, + int32_t lstrip, bool special); + /// @details Convert the provided tokens into text (inverse of llama_tokenize()). + /// @param text The char pointer must be large enough to hold the resulting text. + /// @return Returns the number of chars/bytes on success, no more than text_len_max. + /// @return Returns a negative number on failure - the number of chars/bytes that would have been returned. + /// @param remove_special Allow to remove BOS and EOS tokens if model is configured to do so. + /// @param unparse_special If true, special tokens are rendered in the output. + LLAMA_API int32_t llama_detokenize( + const struct llama_model * model, + const llama_token * tokens, + int32_t n_tokens, + char * text, + int32_t text_len_max, + bool remove_special, + bool unparse_special); + /// Apply chat template. Inspired by hf apply_chat_template() on python. /// Both "model" and "custom_template" are optional, but at least one is required. "custom_template" has higher precedence than "model" /// NOTE: This function does not use a jinja parser. It only support a pre-defined list of template. See more: https://github.com/ggerganov/llama.cpp/wiki/Templates-supported-by-llama_chat_apply_template diff --git a/models/ggml-vocab-bert-bge.gguf.inp b/models/ggml-vocab-bert-bge.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-bert-bge.gguf.inp +++ b/models/ggml-vocab-bert-bge.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-bert-bge.gguf.out b/models/ggml-vocab-bert-bge.gguf.out index e4a76cdb0..a62566ce7 100644 --- a/models/ggml-vocab-bert-bge.gguf.out +++ b/models/ggml-vocab-bert-bge.gguf.out @@ -31,6 +31,7 @@ 1027 1005 3690 7592 1010 1061 1005 2035 999 2129 2024 2017 100 1029 1855 100 100 6207 100 100 14677 23632 22203 1811 1995 + 999 999 999 999 999 999 1017 3943 21211 @@ -40,4 +41,6 @@ 21211 22394 22394 21211 22394 22394 2509 21211 22394 22394 22394 + 12731 2050 19710 + 5860 18117 100 1006 3671 1007 100 1006 3674 7861 29147 2483 9530 16280 23854 1007 100 100 1017 3943 21211 21211 2509 21211 22394 21211 22394 2509 21211 22394 22394 21211 22394 22394 2509 1017 1012 1017 1017 1012 1012 1017 1017 1012 1012 1012 1017 100 1029 1855 100 100 6207 100 100 14677 23632 22203 1811 1995 1011 1011 1011 1011 1011 1011 1027 1027 1027 1027 1027 1027 1027 1192 15290 29754 14150 1192 10260 1181 29755 29436 29741 10260 16856 29747 23925 10325 1005 1005 1005 1005 1005 1005 1036 1036 1036 1036 1036 1036 1036 1000 1000 1000 1000 1012 1012 1012 1012 1012 1012 999 999 999 999 999 999 1029 1029 1029 1029 1029 1029 1045 1005 2310 2042 1005 2409 2002 1005 1055 2045 1010 1005 2128 2017 2469 1029 1005 1049 2025 2469 1045 1005 2222 2191 2009 1010 1005 1040 2017 2066 2070 5572 1029 2057 1005 2310 1037 1005 2222 diff --git a/models/ggml-vocab-command-r.gguf.inp b/models/ggml-vocab-command-r.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-command-r.gguf.inp +++ b/models/ggml-vocab-command-r.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-command-r.gguf.out b/models/ggml-vocab-command-r.gguf.out index cc4277daa..3f6b41888 100644 --- a/models/ggml-vocab-command-r.gguf.out +++ b/models/ggml-vocab-command-r.gguf.out @@ -31,6 +31,7 @@ 206 1857 14 4515 28339 19 1770 14 1954 8 4070 1955 1933 80503 231 5691 12081 13336 2648 29325 14315 24 26 24 27 24 28 24 5123 18372 + 57178 10251 26 26 26 26 26 26 @@ -40,4 +41,6 @@ 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 26 + 42 30719 12584 + 3642 4388 127731 51628 205 57788 18494 97469 126134 206 2226 256 230 1737 18258 16 80503 122 35927 2226 242 112 57462 1737 54457 223165 106230 2096 16 48389 11254 107 255 2226 107 255 228 26 228 26 26 228 26 26 26 228 26 26 26 26 228 26 26 26 26 26 228 26 26 26 26 26 26 228 26 26 26 26 26 26 26 228 26 26 26 26 26 26 26 26 228 26 21 26 228 26 2271 26 228 26 3834 26 182018 230 174833 38111 249 86325 241 38111 245 86325 232 38111 252 38111 123 38111 261 165 24629 38111 261 38111 103 174833 38111 235 188568 231 5691 12081 13336 2648 29325 14315 24 26 24 27 24 28 24 5123 18372 8391 158343 3512 40071 2196 3236 8750 1764 37097 41168 29721 32797 25646 3802 4975 4975 116167 57178 10251 154048 27292 1767 5125 2632 2155 91 2378 1919 1914 2782 19 2155 3354 1933 5470 38 2155 52 2068 5470 1767 4961 3059 1894 19 2155 43 1933 3026 2725 23186 38 2930 14 20676 1671 14 83 51 diff --git a/models/ggml-vocab-deepseek-coder.gguf.inp b/models/ggml-vocab-deepseek-coder.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-deepseek-coder.gguf.inp +++ b/models/ggml-vocab-deepseek-coder.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-deepseek-coder.gguf.out b/models/ggml-vocab-deepseek-coder.gguf.out index 9ccc560d6..52c4111a1 100644 --- a/models/ggml-vocab-deepseek-coder.gguf.out +++ b/models/ggml-vocab-deepseek-coder.gguf.out @@ -31,6 +31,7 @@ 185 405 6 2895 17535 11 320 6 435 0 1717 417 340 12394 233 210 3015 19100 608 9413 2668 16 18 16 19 16 20 16 1393 169 121 239 + 15330 3023 18 18 18 18 18 18 @@ -40,4 +41,6 @@ 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 + 34 155 119 242 64 24297 155 119 216 83 + 1607 2539 185 207 185 185 207 185 185 185 207 12405 459 22758 185 243 185 315 185 251 185 730 185 10047 235 209 334 8760 8 12394 233 114 350 222 10047 221 104 169 116 224 334 4684 3909 992 24330 262 29651 612 8 207 156 237 214 12394 99 234 10047 99 234 207 18 207 18 18 207 18 18 18 207 18 18 18 18 207 18 18 18 18 18 207 18 18 18 18 18 18 207 18 18 18 18 18 18 18 207 18 18 18 18 18 18 18 18 207 18 13 18 207 18 524 18 207 18 1202 18 207 155 239 209 155 239 114 155 239 228 155 240 220 155 239 224 155 240 211 155 239 231 155 239 115 155 239 240 155 240 210 155 239 240 155 239 95 155 239 114 155 239 214 10047 233 210 3015 19100 608 9413 2668 16 18 16 19 16 20 16 1393 169 121 239 18155 374 17194 28 2861 6478 616 2251 14994 31269 4191 6 4686 4686 10252 3358 3358 3409 524 15330 3023 15031 5668 303 6 312 798 651 83 839 362 6 82 741 11 651 1369 340 2037 30 651 44 441 2037 303 6 642 1098 359 11 651 35 340 833 738 10860 30 998 6 10709 245 6 75 43 diff --git a/models/ggml-vocab-deepseek-llm.gguf.inp b/models/ggml-vocab-deepseek-llm.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-deepseek-llm.gguf.inp +++ b/models/ggml-vocab-deepseek-llm.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-deepseek-llm.gguf.out b/models/ggml-vocab-deepseek-llm.gguf.out index fd94b896d..0191b7a11 100644 --- a/models/ggml-vocab-deepseek-llm.gguf.out +++ b/models/ggml-vocab-deepseek-llm.gguf.out @@ -31,6 +31,7 @@ 185 403 6 2906 17464 11 320 6 436 0 1724 418 340 33701 210 3025 19017 612 9407 2681 16 18 16 19 16 20 16 1398 68940 239 + 15278 3033 18 18 18 18 18 18 @@ -40,4 +41,6 @@ 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 + 34 32555 242 64 23708 32555 216 83 + 1763 2550 185 207 185 185 207 185 185 185 207 11969 486 22504 185 243 185 300 185 251 185 663 185 10044 95300 334 8754 8 33701 114 350 222 10044 221 104 46713 334 34732 996 24250 262 80923 8 207 37103 214 12356 99 234 10044 99 234 207 18 207 18 18 207 18 18 18 207 18 18 18 18 207 18 18 18 18 18 207 18 18 18 18 18 18 207 18 18 18 18 18 18 18 207 18 18 18 18 18 18 18 18 207 18 13 18 207 18 526 18 207 18 1204 18 207 71374 209 71374 114 71374 228 155 240 220 71374 224 155 240 211 71374 231 71374 115 71374 240 155 240 210 71374 240 71374 95 71374 114 71374 214 71899 210 3025 19017 612 9407 2681 16 18 16 19 16 20 16 1398 68940 239 78827 55170 76659 620 91754 31116 36804 4885 4885 10897 4390 4390 41047 15278 3033 14986 5675 304 6 313 803 655 33326 362 6 82 745 11 655 1374 340 2049 30 655 44 441 2049 304 6 647 1099 359 11 655 35 340 837 742 10842 30 1003 6 10699 245 6 75 43 diff --git a/models/ggml-vocab-falcon.gguf.inp b/models/ggml-vocab-falcon.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-falcon.gguf.inp +++ b/models/ggml-vocab-falcon.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-falcon.gguf.out b/models/ggml-vocab-falcon.gguf.out index 209b04cda..64a48d97f 100644 --- a/models/ggml-vocab-falcon.gguf.out +++ b/models/ggml-vocab-falcon.gguf.out @@ -31,6 +31,7 @@ 1212 40 18 4932 9856 23 291 18 436 12 1265 362 299 8196 207 204 42 50087 123 2727 20300 32022 133 234 17419 30137 28 7858 181 133 236 + 51520 30 3138 22287 @@ -40,4 +41,6 @@ 22287 22287 30 22287 22287 3138 22287 22287 22287 + 46 19768 239 76 9634 19768 213 95 + 1080 1502 1212 4824 1001 1212 192 204 663 49453 2069 742 561 1501 193 2571 232 206 204 19 11003 20 8196 126 283 219 48778 116 13392 204 19 51831 732 63209 1741 7955 522 20 22438 211 3346 111 231 2571 111 231 204 30 204 3138 204 22287 204 22287 30 204 22287 3138 204 22287 22287 204 22287 22287 30 204 22287 22287 3138 204 30 25 30 204 30 513 30 204 30 951 30 27171 236 206 38154 126 38154 225 167 237 217 38154 221 167 237 208 38154 228 38154 127 38154 237 167 237 207 38154 237 38154 107 38154 126 38154 211 20589 207 204 42 50087 123 2727 20300 32022 133 234 17419 30137 28 7858 181 133 236 204 37057 2228 10666 5052 133 6207 151 215 150 134 5052 133 6279 5052 223 151 216 49679 123 53110 47043 7795 204 7544 7544 7544 8543 8543 17593 3513 3513 12844 51520 17664 4247 295 18 298 650 204 18 95 693 332 18 94 629 23 204 18 1553 299 1310 42 204 18 56 416 1310 295 18 567 717 334 23 204 18 47 299 606 596 6696 42 703 18 16139 241 18 87 55 diff --git a/models/ggml-vocab-gpt-2.gguf.inp b/models/ggml-vocab-gpt-2.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-gpt-2.gguf.inp +++ b/models/ggml-vocab-gpt-2.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-gpt-2.gguf.out b/models/ggml-vocab-gpt-2.gguf.out index 78430f0d3..17a13bdfc 100644 --- a/models/ggml-vocab-gpt-2.gguf.out +++ b/models/ggml-vocab-gpt-2.gguf.out @@ -31,6 +31,7 @@ 198 796 6 6980 15496 11 331 6 439 0 1374 389 345 30325 223 5633 22755 239 46349 111 28839 101 18040 32432 98 43291 1485 1415 24309 25465 171 121 252 + 13896 3228 18 2091 20370 @@ -40,4 +41,6 @@ 24840 20370 24840 24840 24840 2091 20370 + 34 157 119 255 64 16049 157 119 229 83 + 1221 1371 198 220 628 220 628 198 220 197 220 197 197 220 197 198 220 220 198 220 220 220 198 220 220 220 220 198 220 220 220 220 220 198 8582 248 222 357 11265 8 30325 114 447 235 8582 234 104 37929 357 48101 795 13210 271 1673 36686 515 8 14519 227 12520 99 247 8582 99 247 513 4747 23460 513 20370 23460 2091 23460 20370 23460 24840 23460 2091 20370 513 13 18 513 492 18 513 986 18 28053 252 222 157 252 114 157 252 241 157 253 233 157 252 237 157 253 224 157 252 244 157 252 115 157 252 253 157 253 223 157 252 253 157 252 95 157 252 114 157 252 227 47249 223 5633 22755 239 46349 111 28839 101 18040 32432 98 43291 1485 1415 24309 25465 171 121 252 40103 1421 18604 12466 121 16843 141 231 15166 12466 121 16142 12466 239 141 232 30143 140 111 16142 21169 21727 31583 18849 705 39115 6 33153 15506 63 15931 15931 16317 13896 3228 9805 3548 314 1053 587 705 44040 339 338 612 11 705 2200 345 1654 30 705 44 407 1654 314 1183 787 340 11 705 35 345 588 617 8887 30 775 6 26979 257 6 75 43 diff --git a/models/ggml-vocab-llama-bpe.gguf.inp b/models/ggml-vocab-llama-bpe.gguf.inp index 9380bf355..9baf7d77a 100644 --- a/models/ggml-vocab-llama-bpe.gguf.inp +++ b/models/ggml-vocab-llama-bpe.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ @@ -104,5 +110,3 @@ __ggml_vocab_test__ 🚀 (normal) 😶‍🌫️ (multiple emojis concatenated) ✅ 🦙🦙 3 33 333 3333 33333 333333 3333333 33333333 3.3 3..3 3...3 កាន់តែពិសេសអាច😁 ?我想在apple工作1314151天～ ------======= нещо на Български ''''''```````""""......!!!!!!?????? I've been 'told he's there, 'RE you sure? 'M not sure I'll make it, 'D you like some tea? We'Ve a'lL __ggml_vocab_test__ - Việt -__ggml_vocab_test__ diff --git a/models/ggml-vocab-llama-bpe.gguf.out b/models/ggml-vocab-llama-bpe.gguf.out index 1f3607fb6..4b35cf93f 100644 --- a/models/ggml-vocab-llama-bpe.gguf.out +++ b/models/ggml-vocab-llama-bpe.gguf.out @@ -31,6 +31,7 @@ 198 284 6 11639 9906 11 379 65948 0 2650 527 499 27623 223 949 37046 101067 19000 23182 102301 9263 18136 16 36827 21909 + 17523 3001 18 1644 8765 @@ -40,5 +41,6 @@ 8765 8765 18 8765 8765 1644 8765 8765 8765 + 34 91163 101798 + 2624 2402 198 4815 15073 66597 8004 1602 2355 79772 11187 9468 248 222 320 8416 8 27623 114 102470 9468 234 104 31643 320 36773 100166 98634 8 26602 227 11410 99 247 9468 99 247 220 18 220 1644 220 8765 220 8765 18 220 8765 1644 220 8765 8765 220 8765 8765 18 220 8765 8765 1644 220 18 13 18 220 18 497 18 220 18 1131 18 220 21549 222 98629 241 45358 233 21549 237 45358 224 21549 244 21549 115 21549 253 45358 223 21549 253 21549 95 98629 227 76460 223 949 37046 101067 19000 23182 102301 9263 18136 16 36827 21909 56560 54337 19175 102118 13373 64571 34694 3114 112203 80112 3436 106451 14196 14196 74694 3089 3089 29249 17523 3001 27708 7801 358 3077 1027 364 83 820 568 596 1070 11 364 793 499 2771 30 364 44 539 2771 358 3358 1304 433 11 364 35 499 1093 1063 15600 30 1226 6 43712 264 64966 43 - 101798 diff --git a/models/ggml-vocab-llama-spm.gguf.inp b/models/ggml-vocab-llama-spm.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-llama-spm.gguf.inp +++ b/models/ggml-vocab-llama-spm.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-llama-spm.gguf.out b/models/ggml-vocab-llama-spm.gguf.out index 9c3327cb5..93aacf8ba 100644 --- a/models/ggml-vocab-llama-spm.gguf.out +++ b/models/ggml-vocab-llama-spm.gguf.out @@ -31,6 +31,7 @@ 29871 13 353 525 3152 15043 29892 343 29915 497 29991 1128 526 366 29871 243 162 155 132 1577 30672 31522 30505 11548 31041 30732 29896 29941 29896 29946 29896 29945 29896 30408 30739 + 1738 6824 21004 29871 29941 29871 29941 29941 29871 29941 29941 29941 @@ -40,4 +41,6 @@ 29871 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29941 + 315 228 190 176 29874 10630 30529 29873 + 29871 2313 3163 29871 13 29871 13 13 29871 13 13 13 29871 12 29871 12 12 29871 12 13 259 13 1678 13 268 13 418 13 243 162 157 131 313 8945 29897 29871 243 162 155 185 30722 243 162 143 174 30598 313 20787 953 3848 275 16125 630 29897 29871 31681 29871 243 162 169 156 243 162 169 156 29871 29941 29871 29941 29941 29871 29941 29941 29941 29871 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29871 29941 29889 29941 29871 29941 636 29941 29871 29941 856 29941 29871 31849 31324 31934 228 162 142 228 161 146 228 162 133 228 161 153 228 161 186 31708 228 162 132 31708 228 161 165 31324 228 161 136 243 162 155 132 1577 30672 31522 30505 11548 31041 30732 29896 29941 29896 29946 29896 29945 29896 30408 30739 448 23648 2751 25512 1538 4851 665 1386 29713 1305 14550 4907 11120 16159 16159 16159 15945 15945 3045 636 6824 6824 6824 8773 8773 8773 306 29915 345 1063 525 29873 1025 540 29915 29879 727 29892 525 1525 366 1854 29973 525 29924 451 1854 306 29915 645 1207 372 29892 525 29928 366 763 777 23429 29973 1334 29915 29963 29872 263 29915 29880 29931 diff --git a/models/ggml-vocab-mpt.gguf.inp b/models/ggml-vocab-mpt.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-mpt.gguf.inp +++ b/models/ggml-vocab-mpt.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-mpt.gguf.out b/models/ggml-vocab-mpt.gguf.out index d8d0fe909..372c751bf 100644 --- a/models/ggml-vocab-mpt.gguf.out +++ b/models/ggml-vocab-mpt.gguf.out @@ -31,6 +31,7 @@ 187 426 8 8685 12092 13 340 8 455 2 1359 403 368 49042 212 3736 15367 41197 13610 19934 41869 21275 1012 1047 18795 40120 20422 241 + 18963 4672 20 1610 20084 @@ -40,4 +41,6 @@ 26409 20084 26409 26409 26409 1610 20084 + 36 6829 244 66 17721 35177 85 + 1262 2196 586 1744 33525 186 209 623 28910 187 50276 187 50275 187 50274 187 50273 187 14931 237 211 313 6320 10 49042 116 325 224 14931 223 106 171 118 226 313 34263 802 13511 261 32147 456 10 3384 239 216 22692 101 236 14931 101 236 495 5922 30057 495 20084 495 26409 30057 20084 495 26409 1610 495 26409 20084 495 15 20 495 537 20 495 1051 20 209 18081 211 18081 116 18081 230 39936 222 18081 226 39936 213 18081 233 18081 117 18081 242 39936 212 18081 242 18081 97 18081 116 18081 216 14931 235 212 3736 15367 41197 13610 19934 41869 21275 1012 1047 18795 40120 20422 241 16081 6877 12880 11514 1068 8713 38177 13396 3415 9925 12559 10453 1389 42011 35033 34842 11202 9739 9739 33021 18963 4672 25561 8220 309 1849 644 686 42618 344 434 627 13 686 1848 368 2119 32 686 46 417 2119 309 1833 1056 352 13 686 37 368 751 690 10331 32 844 8 31516 247 8 77 45 diff --git a/models/ggml-vocab-phi-3.gguf.inp b/models/ggml-vocab-phi-3.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-phi-3.gguf.inp +++ b/models/ggml-vocab-phi-3.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-phi-3.gguf.out b/models/ggml-vocab-phi-3.gguf.out index 9c3327cb5..93aacf8ba 100644 --- a/models/ggml-vocab-phi-3.gguf.out +++ b/models/ggml-vocab-phi-3.gguf.out @@ -31,6 +31,7 @@ 29871 13 353 525 3152 15043 29892 343 29915 497 29991 1128 526 366 29871 243 162 155 132 1577 30672 31522 30505 11548 31041 30732 29896 29941 29896 29946 29896 29945 29896 30408 30739 + 1738 6824 21004 29871 29941 29871 29941 29941 29871 29941 29941 29941 @@ -40,4 +41,6 @@ 29871 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29941 + 315 228 190 176 29874 10630 30529 29873 + 29871 2313 3163 29871 13 29871 13 13 29871 13 13 13 29871 12 29871 12 12 29871 12 13 259 13 1678 13 268 13 418 13 243 162 157 131 313 8945 29897 29871 243 162 155 185 30722 243 162 143 174 30598 313 20787 953 3848 275 16125 630 29897 29871 31681 29871 243 162 169 156 243 162 169 156 29871 29941 29871 29941 29941 29871 29941 29941 29941 29871 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29871 29941 29941 29941 29941 29941 29941 29941 29941 29871 29941 29889 29941 29871 29941 636 29941 29871 29941 856 29941 29871 31849 31324 31934 228 162 142 228 161 146 228 162 133 228 161 153 228 161 186 31708 228 162 132 31708 228 161 165 31324 228 161 136 243 162 155 132 1577 30672 31522 30505 11548 31041 30732 29896 29941 29896 29946 29896 29945 29896 30408 30739 448 23648 2751 25512 1538 4851 665 1386 29713 1305 14550 4907 11120 16159 16159 16159 15945 15945 3045 636 6824 6824 6824 8773 8773 8773 306 29915 345 1063 525 29873 1025 540 29915 29879 727 29892 525 1525 366 1854 29973 525 29924 451 1854 306 29915 645 1207 372 29892 525 29928 366 763 777 23429 29973 1334 29915 29963 29872 263 29915 29880 29931 diff --git a/models/ggml-vocab-qwen2.gguf.inp b/models/ggml-vocab-qwen2.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-qwen2.gguf.inp +++ b/models/ggml-vocab-qwen2.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-qwen2.gguf.out b/models/ggml-vocab-qwen2.gguf.out index 401a510e8..18b4b45cd 100644 --- a/models/ggml-vocab-qwen2.gguf.out +++ b/models/ggml-vocab-qwen2.gguf.out @@ -31,6 +31,7 @@ 198 284 6 11385 9707 11 379 64848 0 2585 525 498 26525 223 937 104100 18493 22377 99257 16 18 16 19 16 20 16 35727 21216 + 17085 2928 18 18 18 18 18 18 @@ -40,4 +41,6 @@ 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 18 + 34 90063 128324 + 2560 2347 198 4710 14731 65497 7847 1572 2303 78672 10947 145836 320 8252 8 26525 114 378 235 149921 30543 320 35673 99066 97534 8 25521 227 11162 99 247 149955 220 18 220 18 18 220 18 18 18 220 18 18 18 18 220 18 18 18 18 18 220 18 18 18 18 18 18 220 18 18 18 18 18 18 18 220 18 18 18 18 18 18 18 18 220 18 13 18 220 18 496 18 220 18 1112 18 220 146394 97529 241 44258 233 146568 44258 224 147603 20879 115 146280 44258 223 146280 147272 97529 227 144534 937 104100 18493 22377 99257 16 18 16 19 16 20 16 35727 21216 55460 53237 18658 14144 1456 13073 63471 33594 3038 133178 79012 3355 4605 4605 13874 13874 73594 3014 3014 28149 17085 2928 26610 7646 358 3003 1012 364 83 813 566 594 1052 11 364 787 498 2704 30 364 44 537 2704 358 3278 1281 432 11 364 35 498 1075 1045 15243 30 1205 6 42612 264 63866 43 diff --git a/models/ggml-vocab-refact.gguf.inp b/models/ggml-vocab-refact.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-refact.gguf.inp +++ b/models/ggml-vocab-refact.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-refact.gguf.out b/models/ggml-vocab-refact.gguf.out index 06b15c090..63d8305c3 100644 --- a/models/ggml-vocab-refact.gguf.out +++ b/models/ggml-vocab-refact.gguf.out @@ -31,6 +31,7 @@ 203 280 25 34666 8279 30 533 25 464 19 4971 884 844 18458 228 1018 4982 13368 2909 9513 17827 35 37 35 38 35 39 35 11873 47838 + 9163 3202 37 37 37 37 37 37 @@ -40,4 +41,6 @@ 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 37 + 53 33934 83 33217 17102 102 + 1214 12258 334 719 8878 202 10885 4222 16104 28570 203 3807 253 227 308 4382 27 18458 133 46113 44967 123 13868 308 12565 19775 33071 40824 733 27 41889 5945 118 252 3807 118 252 225 37 225 37 37 225 37 37 37 225 37 37 37 37 225 37 37 37 37 37 225 37 37 37 37 37 37 225 37 37 37 37 37 37 37 225 37 37 37 37 37 37 37 37 225 37 32 37 225 37 497 37 225 37 1179 37 225 14574 227 14574 133 14574 246 30457 238 14574 242 30457 229 14574 249 14574 134 14574 258 30457 228 14574 258 14574 114 14574 133 14574 232 36628 228 1018 4982 13368 2909 9513 17827 35 37 35 38 35 39 35 11873 47838 20921 16623 13028 8372 1039 9446 40242 13852 2053 8949 12531 1520 10700 5881 9592 13299 914 31753 31359 9163 3202 35472 10397 439 4763 2583 330 102 1455 938 1182 2017 30 330 613 844 3654 49 330 63 646 3654 439 4621 1930 561 30 330 54 844 2124 1629 35993 49 2688 25 7709 312 25 94 62 diff --git a/models/ggml-vocab-starcoder.gguf.inp b/models/ggml-vocab-starcoder.gguf.inp index 0a89107c6..9baf7d77a 100644 --- a/models/ggml-vocab-starcoder.gguf.inp +++ b/models/ggml-vocab-starcoder.gguf.inp @@ -73,6 +73,8 @@ __ggml_vocab_test__ __ggml_vocab_test__ Hello, y'all! How are you 😁 ?我想在apple工作1314151天～ __ggml_vocab_test__ +!!!!!! +__ggml_vocab_test__ 3 __ggml_vocab_test__ 33 @@ -91,6 +93,10 @@ __ggml_vocab_test__ __ggml_vocab_test__ 333333333 __ggml_vocab_test__ +Cửa Việt +__ggml_vocab_test__ + discards +__ggml_vocab_test__ diff --git a/models/ggml-vocab-starcoder.gguf.out b/models/ggml-vocab-starcoder.gguf.out index ccb55c7fe..87e2465d3 100644 --- a/models/ggml-vocab-starcoder.gguf.out +++ b/models/ggml-vocab-starcoder.gguf.out @@ -31,6 +31,7 @@ 222 299 44 34719 8302 49 553 44 483 38 4998 904 863 18445 247 1037 4995 13379 2924 9515 17823 54 56 54 57 54 58 54 11904 47892 + 9221 3226 56 56 56 56 56 56 @@ -40,4 +41,6 @@ 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 56 + 72 34269 102 33245 17234 121 + 1236 12266 353 736 8886 221 10883 4238 16101 28540 222 3822 272 246 327 4434 46 18445 152 46030 45022 142 13878 327 12585 19884 33773 40920 751 46 41839 5954 137 271 3822 137 271 244 56 244 56 56 244 56 56 56 244 56 56 56 56 244 56 56 56 56 56 244 56 56 56 56 56 56 244 56 56 56 56 56 56 56 244 56 56 56 56 56 56 56 56 244 56 51 56 244 56 516 56 244 56 1198 56 244 14566 246 14566 152 14566 265 30428 257 14566 261 30428 248 14566 268 14566 153 14566 277 30428 247 14566 277 14566 133 14566 152 14566 251 36570 247 1037 4995 13379 2924 9515 17823 54 56 54 57 54 58 54 11904 47892 20895 16625 13047 8389 1059 9504 40216 13858 2073 8983 12571 1539 10721 5918 9643 13298 932 31723 31330 9221 3226 35426 10400 457 4783 2602 349 121 1477 957 1200 2038 49 349 632 863 3673 68 349 82 666 3673 457 4650 1949 580 49 349 73 863 2144 1649 35941 68 2726 44 7728 331 44 113 81 diff --git a/poetry.lock b/poetry.lock new file mode 100644 index 000000000..eb6baa6c7 --- /dev/null +++ b/poetry.lock @@ -0,0 +1,1197 @@ +# This file is automatically @generated by Poetry 1.7.1 and should not be changed by hand. + +[[package]] +name = "atomicwrites" +version = "1.4.1" +description = "Atomic file writes." +optional = false +python-versions = ">=2.7, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*" +files = [ + {file = "atomicwrites-1.4.1.tar.gz", hash = "sha256:81b2c9071a49367a7f770170e5eec8cb66567cfbbc8c73d20ce5ca4a8d71cf11"}, +] + +[[package]] +name = "attrs" +version = "23.2.0" +description = "Classes Without Boilerplate" +optional = false +python-versions = ">=3.7" +files = [ + {file = "attrs-23.2.0-py3-none-any.whl", hash = "sha256:99b87a485a5820b23b879f04c2305b44b951b502fd64be915879d77a7e8fc6f1"}, + {file = "attrs-23.2.0.tar.gz", hash = "sha256:935dc3b529c262f6cf76e50877d35a4bd3c1de194fd41f47a2b7ae8f19971f30"}, +] + +[package.extras] +cov = ["attrs[tests]", "coverage[toml] (>=5.3)"] +dev = ["attrs[tests]", "pre-commit"] +docs = ["furo", "myst-parser", "sphinx", "sphinx-notfound-page", "sphinxcontrib-towncrier", "towncrier", "zope-interface"] +tests = ["attrs[tests-no-zope]", "zope-interface"] +tests-mypy = ["mypy (>=1.6)", "pytest-mypy-plugins"] +tests-no-zope = ["attrs[tests-mypy]", "cloudpickle", "hypothesis", "pympler", "pytest (>=4.3.0)", "pytest-xdist[psutil]"] + +[[package]] +name = "certifi" +version = "2024.2.2" +description = "Python package for providing Mozilla's CA Bundle." +optional = false +python-versions = ">=3.6" +files = [ + {file = "certifi-2024.2.2-py3-none-any.whl", hash = "sha256:dc383c07b76109f368f6106eee2b593b04a011ea4d55f652c6ca24a754d1cdd1"}, + {file = "certifi-2024.2.2.tar.gz", hash = "sha256:0569859f95fc761b18b45ef421b1290a0f65f147e92a1e5eb3e635f9a5e4e66f"}, +] + +[[package]] +name = "charset-normalizer" +version = "3.3.2" +description = "The Real First Universal Charset Detector. Open, modern and actively maintained alternative to Chardet." +optional = false +python-versions = ">=3.7.0" +files = [ + {file = "charset-normalizer-3.3.2.tar.gz", hash = "sha256:f30c3cb33b24454a82faecaf01b19c18562b1e89558fb6c56de4d9118a032fd5"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:25baf083bf6f6b341f4121c2f3c548875ee6f5339300e08be3f2b2ba1721cdd3"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:06435b539f889b1f6f4ac1758871aae42dc3a8c0e24ac9e60c2384973ad73027"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:9063e24fdb1e498ab71cb7419e24622516c4a04476b17a2dab57e8baa30d6e03"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6897af51655e3691ff853668779c7bad41579facacf5fd7253b0133308cf000d"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1d3193f4a680c64b4b6a9115943538edb896edc190f0b222e73761716519268e"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:cd70574b12bb8a4d2aaa0094515df2463cb429d8536cfb6c7ce983246983e5a6"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8465322196c8b4d7ab6d1e049e4c5cb460d0394da4a27d23cc242fbf0034b6b5"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:a9a8e9031d613fd2009c182b69c7b2c1ef8239a0efb1df3f7c8da66d5dd3d537"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:beb58fe5cdb101e3a055192ac291b7a21e3b7ef4f67fa1d74e331a7f2124341c"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:e06ed3eb3218bc64786f7db41917d4e686cc4856944f53d5bdf83a6884432e12"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-musllinux_1_1_ppc64le.whl", hash = "sha256:2e81c7b9c8979ce92ed306c249d46894776a909505d8f5a4ba55b14206e3222f"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-musllinux_1_1_s390x.whl", hash = "sha256:572c3763a264ba47b3cf708a44ce965d98555f618ca42c926a9c1616d8f34269"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:fd1abc0d89e30cc4e02e4064dc67fcc51bd941eb395c502aac3ec19fab46b519"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-win32.whl", hash = "sha256:3d47fa203a7bd9c5b6cee4736ee84ca03b8ef23193c0d1ca99b5089f72645c73"}, + {file = "charset_normalizer-3.3.2-cp310-cp310-win_amd64.whl", hash = "sha256:10955842570876604d404661fbccbc9c7e684caf432c09c715ec38fbae45ae09"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:802fe99cca7457642125a8a88a084cef28ff0cf9407060f7b93dca5aa25480db"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:573f6eac48f4769d667c4442081b1794f52919e7edada77495aaed9236d13a96"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:549a3a73da901d5bc3ce8d24e0600d1fa85524c10287f6004fbab87672bf3e1e"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f27273b60488abe721a075bcca6d7f3964f9f6f067c8c4c605743023d7d3944f"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:1ceae2f17a9c33cb48e3263960dc5fc8005351ee19db217e9b1bb15d28c02574"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:65f6f63034100ead094b8744b3b97965785388f308a64cf8d7c34f2f2e5be0c4"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:753f10e867343b4511128c6ed8c82f7bec3bd026875576dfd88483c5c73b2fd8"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:4a78b2b446bd7c934f5dcedc588903fb2f5eec172f3d29e52a9096a43722adfc"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:e537484df0d8f426ce2afb2d0f8e1c3d0b114b83f8850e5f2fbea0e797bd82ae"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:eb6904c354526e758fda7167b33005998fb68c46fbc10e013ca97f21ca5c8887"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-musllinux_1_1_ppc64le.whl", hash = "sha256:deb6be0ac38ece9ba87dea880e438f25ca3eddfac8b002a2ec3d9183a454e8ae"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-musllinux_1_1_s390x.whl", hash = "sha256:4ab2fe47fae9e0f9dee8c04187ce5d09f48eabe611be8259444906793ab7cbce"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:80402cd6ee291dcb72644d6eac93785fe2c8b9cb30893c1af5b8fdd753b9d40f"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-win32.whl", hash = "sha256:7cd13a2e3ddeed6913a65e66e94b51d80a041145a026c27e6bb76c31a853c6ab"}, + {file = "charset_normalizer-3.3.2-cp311-cp311-win_amd64.whl", hash = "sha256:663946639d296df6a2bb2aa51b60a2454ca1cb29835324c640dafb5ff2131a77"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:0b2b64d2bb6d3fb9112bafa732def486049e63de9618b5843bcdd081d8144cd8"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:ddbb2551d7e0102e7252db79ba445cdab71b26640817ab1e3e3648dad515003b"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:55086ee1064215781fff39a1af09518bc9255b50d6333f2e4c74ca09fac6a8f6"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8f4a014bc36d3c57402e2977dada34f9c12300af536839dc38c0beab8878f38a"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a10af20b82360ab00827f916a6058451b723b4e65030c5a18577c8b2de5b3389"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:8d756e44e94489e49571086ef83b2bb8ce311e730092d2c34ca8f7d925cb20aa"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:90d558489962fd4918143277a773316e56c72da56ec7aa3dc3dbbe20fdfed15b"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:6ac7ffc7ad6d040517be39eb591cac5ff87416c2537df6ba3cba3bae290c0fed"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:7ed9e526742851e8d5cc9e6cf41427dfc6068d4f5a3bb03659444b4cabf6bc26"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-musllinux_1_1_i686.whl", hash = "sha256:8bdb58ff7ba23002a4c5808d608e4e6c687175724f54a5dade5fa8c67b604e4d"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-musllinux_1_1_ppc64le.whl", hash = "sha256:6b3251890fff30ee142c44144871185dbe13b11bab478a88887a639655be1068"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-musllinux_1_1_s390x.whl", hash = "sha256:b4a23f61ce87adf89be746c8a8974fe1c823c891d8f86eb218bb957c924bb143"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:efcb3f6676480691518c177e3b465bcddf57cea040302f9f4e6e191af91174d4"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-win32.whl", hash = "sha256:d965bba47ddeec8cd560687584e88cf699fd28f192ceb452d1d7ee807c5597b7"}, + {file = "charset_normalizer-3.3.2-cp312-cp312-win_amd64.whl", hash = "sha256:96b02a3dc4381e5494fad39be677abcb5e6634bf7b4fa83a6dd3112607547001"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:95f2a5796329323b8f0512e09dbb7a1860c46a39da62ecb2324f116fa8fdc85c"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c002b4ffc0be611f0d9da932eb0f704fe2602a9a949d1f738e4c34c75b0863d5"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:a981a536974bbc7a512cf44ed14938cf01030a99e9b3a06dd59578882f06f985"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3287761bc4ee9e33561a7e058c72ac0938c4f57fe49a09eae428fd88aafe7bb6"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:42cb296636fcc8b0644486d15c12376cb9fa75443e00fb25de0b8602e64c1714"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:0a55554a2fa0d408816b3b5cedf0045f4b8e1a6065aec45849de2d6f3f8e9786"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:c083af607d2515612056a31f0a8d9e0fcb5876b7bfc0abad3ecd275bc4ebc2d5"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:87d1351268731db79e0f8e745d92493ee2841c974128ef629dc518b937d9194c"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-musllinux_1_1_ppc64le.whl", hash = "sha256:bd8f7df7d12c2db9fab40bdd87a7c09b1530128315d047a086fa3ae3435cb3a8"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-musllinux_1_1_s390x.whl", hash = "sha256:c180f51afb394e165eafe4ac2936a14bee3eb10debc9d9e4db8958fe36afe711"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:8c622a5fe39a48f78944a87d4fb8a53ee07344641b0562c540d840748571b811"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-win32.whl", hash = "sha256:db364eca23f876da6f9e16c9da0df51aa4f104a972735574842618b8c6d999d4"}, + {file = "charset_normalizer-3.3.2-cp37-cp37m-win_amd64.whl", hash = "sha256:86216b5cee4b06df986d214f664305142d9c76df9b6512be2738aa72a2048f99"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:6463effa3186ea09411d50efc7d85360b38d5f09b870c48e4600f63af490e56a"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:6c4caeef8fa63d06bd437cd4bdcf3ffefe6738fb1b25951440d80dc7df8c03ac"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:37e55c8e51c236f95b033f6fb391d7d7970ba5fe7ff453dad675e88cf303377a"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:fb69256e180cb6c8a894fee62b3afebae785babc1ee98b81cdf68bbca1987f33"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:ae5f4161f18c61806f411a13b0310bea87f987c7d2ecdbdaad0e94eb2e404238"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:b2b0a0c0517616b6869869f8c581d4eb2dd83a4d79e0ebcb7d373ef9956aeb0a"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:45485e01ff4d3630ec0d9617310448a8702f70e9c01906b0d0118bdf9d124cf2"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:eb00ed941194665c332bf8e078baf037d6c35d7c4f3102ea2d4f16ca94a26dc8"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:2127566c664442652f024c837091890cb1942c30937add288223dc895793f898"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:a50aebfa173e157099939b17f18600f72f84eed3049e743b68ad15bd69b6bf99"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-musllinux_1_1_ppc64le.whl", hash = "sha256:4d0d1650369165a14e14e1e47b372cfcb31d6ab44e6e33cb2d4e57265290044d"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-musllinux_1_1_s390x.whl", hash = "sha256:923c0c831b7cfcb071580d3f46c4baf50f174be571576556269530f4bbd79d04"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:06a81e93cd441c56a9b65d8e1d043daeb97a3d0856d177d5c90ba85acb3db087"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-win32.whl", hash = "sha256:6ef1d82a3af9d3eecdba2321dc1b3c238245d890843e040e41e470ffa64c3e25"}, + {file = "charset_normalizer-3.3.2-cp38-cp38-win_amd64.whl", hash = "sha256:eb8821e09e916165e160797a6c17edda0679379a4be5c716c260e836e122f54b"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:c235ebd9baae02f1b77bcea61bce332cb4331dc3617d254df3323aa01ab47bd4"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:5b4c145409bef602a690e7cfad0a15a55c13320ff7a3ad7ca59c13bb8ba4d45d"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:68d1f8a9e9e37c1223b656399be5d6b448dea850bed7d0f87a8311f1ff3dabb0"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:22afcb9f253dac0696b5a4be4a1c0f8762f8239e21b99680099abd9b2b1b2269"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:e27ad930a842b4c5eb8ac0016b0a54f5aebbe679340c26101df33424142c143c"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:1f79682fbe303db92bc2b1136016a38a42e835d932bab5b3b1bfcfbf0640e519"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b261ccdec7821281dade748d088bb6e9b69e6d15b30652b74cbbac25e280b796"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:122c7fa62b130ed55f8f285bfd56d5f4b4a5b503609d181f9ad85e55c89f4185"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:d0eccceffcb53201b5bfebb52600a5fb483a20b61da9dbc885f8b103cbe7598c"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:9f96df6923e21816da7e0ad3fd47dd8f94b2a5ce594e00677c0013018b813458"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-musllinux_1_1_ppc64le.whl", hash = "sha256:7f04c839ed0b6b98b1a7501a002144b76c18fb1c1850c8b98d458ac269e26ed2"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-musllinux_1_1_s390x.whl", hash = "sha256:34d1c8da1e78d2e001f363791c98a272bb734000fcef47a491c1e3b0505657a8"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:ff8fa367d09b717b2a17a052544193ad76cd49979c805768879cb63d9ca50561"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-win32.whl", hash = "sha256:aed38f6e4fb3f5d6bf81bfa990a07806be9d83cf7bacef998ab1a9bd660a581f"}, + {file = "charset_normalizer-3.3.2-cp39-cp39-win_amd64.whl", hash = "sha256:b01b88d45a6fcb69667cd6d2f7a9aeb4bf53760d7fc536bf679ec94fe9f3ff3d"}, + {file = "charset_normalizer-3.3.2-py3-none-any.whl", hash = "sha256:3e4d1f6587322d2788836a99c69062fbb091331ec940e02d12d179c1d53e25fc"}, +] + +[[package]] +name = "colorama" +version = "0.4.6" +description = "Cross-platform colored terminal text." +optional = false +python-versions = "!=3.0.*,!=3.1.*,!=3.2.*,!=3.3.*,!=3.4.*,!=3.5.*,!=3.6.*,>=2.7" +files = [ + {file = "colorama-0.4.6-py2.py3-none-any.whl", hash = "sha256:4f1d9991f5acc0ca119f9d443620b77f9d6b33703e51011c16baf57afb285fc6"}, + {file = "colorama-0.4.6.tar.gz", hash = "sha256:08695f5cb7ed6e0531a20572697297273c47b8cae5a63ffc6d6ed5c201be6e44"}, +] + +[[package]] +name = "filelock" +version = "3.13.1" +description = "A platform independent file lock." +optional = false +python-versions = ">=3.8" +files = [ + {file = "filelock-3.13.1-py3-none-any.whl", hash = "sha256:57dbda9b35157b05fb3e58ee91448612eb674172fab98ee235ccb0b5bee19a1c"}, + {file = "filelock-3.13.1.tar.gz", hash = "sha256:521f5f56c50f8426f5e03ad3b281b490a87ef15bc6c526f168290f0c7148d44e"}, +] + +[package.extras] +docs = ["furo (>=2023.9.10)", "sphinx (>=7.2.6)", "sphinx-autodoc-typehints (>=1.24)"] +testing = ["covdefaults (>=2.3)", "coverage (>=7.3.2)", "diff-cover (>=8)", "pytest (>=7.4.3)", "pytest-cov (>=4.1)", "pytest-mock (>=3.12)", "pytest-timeout (>=2.2)"] +typing = ["typing-extensions (>=4.8)"] + +[[package]] +name = "fsspec" +version = "2024.2.0" +description = "File-system specification" +optional = false +python-versions = ">=3.8" +files = [ + {file = "fsspec-2024.2.0-py3-none-any.whl", hash = "sha256:817f969556fa5916bc682e02ca2045f96ff7f586d45110fcb76022063ad2c7d8"}, + {file = "fsspec-2024.2.0.tar.gz", hash = "sha256:b6ad1a679f760dda52b1168c859d01b7b80648ea6f7f7c7f5a8a91dc3f3ecb84"}, +] + +[package.extras] +abfs = ["adlfs"] +adl = ["adlfs"] +arrow = ["pyarrow (>=1)"] +dask = ["dask", "distributed"] +devel = ["pytest", "pytest-cov"] +dropbox = ["dropbox", "dropboxdrivefs", "requests"] +full = ["adlfs", "aiohttp (!=4.0.0a0,!=4.0.0a1)", "dask", "distributed", "dropbox", "dropboxdrivefs", "fusepy", "gcsfs", "libarchive-c", "ocifs", "panel", "paramiko", "pyarrow (>=1)", "pygit2", "requests", "s3fs", "smbprotocol", "tqdm"] +fuse = ["fusepy"] +gcs = ["gcsfs"] +git = ["pygit2"] +github = ["requests"] +gs = ["gcsfs"] +gui = ["panel"] +hdfs = ["pyarrow (>=1)"] +http = ["aiohttp (!=4.0.0a0,!=4.0.0a1)"] +libarchive = ["libarchive-c"] +oci = ["ocifs"] +s3 = ["s3fs"] +sftp = ["paramiko"] +smb = ["smbprotocol"] +ssh = ["paramiko"] +tqdm = ["tqdm"] + +[[package]] +name = "gguf" +version = "0.7.0" +description = "Read and write ML models in GGUF for GGML" +optional = false +python-versions = ">=3.8" +files = [] +develop = false + +[package.dependencies] +numpy = ">=1.17" + +[package.source] +type = "directory" +url = "gguf-py" + +[[package]] +name = "huggingface-hub" +version = "0.20.3" +description = "Client library to download and publish models, datasets and other repos on the huggingface.co hub" +optional = false +python-versions = ">=3.8.0" +files = [ + {file = "huggingface_hub-0.20.3-py3-none-any.whl", hash = "sha256:d988ae4f00d3e307b0c80c6a05ca6dbb7edba8bba3079f74cda7d9c2e562a7b6"}, + {file = "huggingface_hub-0.20.3.tar.gz", hash = "sha256:94e7f8e074475fbc67d6a71957b678e1b4a74ff1b64a644fd6cbb83da962d05d"}, +] + +[package.dependencies] +filelock = "*" +fsspec = ">=2023.5.0" +packaging = ">=20.9" +pyyaml = ">=5.1" +requests = "*" +tqdm = ">=4.42.1" +typing-extensions = ">=3.7.4.3" + +[package.extras] +all = ["InquirerPy (==0.3.4)", "Jinja2", "Pillow", "aiohttp", "gradio", "jedi", "mypy (==1.5.1)", "numpy", "pydantic (>1.1,<2.0)", "pydantic (>1.1,<3.0)", "pytest", "pytest-asyncio", "pytest-cov", "pytest-env", "pytest-rerunfailures", "pytest-vcr", "pytest-xdist", "ruff (>=0.1.3)", "soundfile", "types-PyYAML", "types-requests", "types-simplejson", "types-toml", "types-tqdm", "types-urllib3", "typing-extensions (>=4.8.0)", "urllib3 (<2.0)"] +cli = ["InquirerPy (==0.3.4)"] +dev = ["InquirerPy (==0.3.4)", "Jinja2", "Pillow", "aiohttp", "gradio", "jedi", "mypy (==1.5.1)", "numpy", "pydantic (>1.1,<2.0)", "pydantic (>1.1,<3.0)", "pytest", "pytest-asyncio", "pytest-cov", "pytest-env", "pytest-rerunfailures", "pytest-vcr", "pytest-xdist", "ruff (>=0.1.3)", "soundfile", "types-PyYAML", "types-requests", "types-simplejson", "types-toml", "types-tqdm", "types-urllib3", "typing-extensions (>=4.8.0)", "urllib3 (<2.0)"] +fastai = ["fastai (>=2.4)", "fastcore (>=1.3.27)", "toml"] +inference = ["aiohttp", "pydantic (>1.1,<2.0)", "pydantic (>1.1,<3.0)"] +quality = ["mypy (==1.5.1)", "ruff (>=0.1.3)"] +tensorflow = ["graphviz", "pydot", "tensorflow"] +testing = ["InquirerPy (==0.3.4)", "Jinja2", "Pillow", "aiohttp", "gradio", "jedi", "numpy", "pydantic (>1.1,<2.0)", "pydantic (>1.1,<3.0)", "pytest", "pytest-asyncio", "pytest-cov", "pytest-env", "pytest-rerunfailures", "pytest-vcr", "pytest-xdist", "soundfile", "urllib3 (<2.0)"] +torch = ["torch"] +typing = ["types-PyYAML", "types-requests", "types-simplejson", "types-toml", "types-tqdm", "types-urllib3", "typing-extensions (>=4.8.0)"] + +[[package]] +name = "idna" +version = "3.6" +description = "Internationalized Domain Names in Applications (IDNA)" +optional = false +python-versions = ">=3.5" +files = [ + {file = "idna-3.6-py3-none-any.whl", hash = "sha256:c05567e9c24a6b9faaa835c4821bad0590fbb9d5779e7caa6e1cc4978e7eb24f"}, + {file = "idna-3.6.tar.gz", hash = "sha256:9ecdbbd083b06798ae1e86adcbfe8ab1479cf864e4ee30fe4e46a003d12491ca"}, +] + +[[package]] +name = "jinja2" +version = "3.1.3" +description = "A very fast and expressive template engine." +optional = false +python-versions = ">=3.7" +files = [ + {file = "Jinja2-3.1.3-py3-none-any.whl", hash = "sha256:7d6d50dd97d52cbc355597bd845fabfbac3f551e1f99619e39a35ce8c370b5fa"}, + {file = "Jinja2-3.1.3.tar.gz", hash = "sha256:ac8bd6544d4bb2c9792bf3a159e80bba8fda7f07e81bc3aed565432d5925ba90"}, +] + +[package.dependencies] +MarkupSafe = ">=2.0" + +[package.extras] +i18n = ["Babel (>=2.7)"] + +[[package]] +name = "markupsafe" +version = "2.1.5" +description = "Safely add untrusted strings to HTML/XML markup." +optional = false +python-versions = ">=3.7" +files = [ + {file = "MarkupSafe-2.1.5-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:a17a92de5231666cfbe003f0e4b9b3a7ae3afb1ec2845aadc2bacc93ff85febc"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:72b6be590cc35924b02c78ef34b467da4ba07e4e0f0454a2c5907f473fc50ce5"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:e61659ba32cf2cf1481e575d0462554625196a1f2fc06a1c777d3f48e8865d46"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:2174c595a0d73a3080ca3257b40096db99799265e1c27cc5a610743acd86d62f"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ae2ad8ae6ebee9d2d94b17fb62763125f3f374c25618198f40cbb8b525411900"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:075202fa5b72c86ad32dc7d0b56024ebdbcf2048c0ba09f1cde31bfdd57bcfff"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:598e3276b64aff0e7b3451b72e94fa3c238d452e7ddcd893c3ab324717456bad"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:fce659a462a1be54d2ffcacea5e3ba2d74daa74f30f5f143fe0c58636e355fdd"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-win32.whl", hash = "sha256:d9fad5155d72433c921b782e58892377c44bd6252b5af2f67f16b194987338a4"}, + {file = "MarkupSafe-2.1.5-cp310-cp310-win_amd64.whl", hash = "sha256:bf50cd79a75d181c9181df03572cdce0fbb75cc353bc350712073108cba98de5"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:629ddd2ca402ae6dbedfceeba9c46d5f7b2a61d9749597d4307f943ef198fc1f"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:5b7b716f97b52c5a14bffdf688f971b2d5ef4029127f1ad7a513973cfd818df2"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6ec585f69cec0aa07d945b20805be741395e28ac1627333b1c5b0105962ffced"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b91c037585eba9095565a3556f611e3cbfaa42ca1e865f7b8015fe5c7336d5a5"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:7502934a33b54030eaf1194c21c692a534196063db72176b0c4028e140f8f32c"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:0e397ac966fdf721b2c528cf028494e86172b4feba51d65f81ffd65c63798f3f"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:c061bb86a71b42465156a3ee7bd58c8c2ceacdbeb95d05a99893e08b8467359a"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:3a57fdd7ce31c7ff06cdfbf31dafa96cc533c21e443d57f5b1ecc6cdc668ec7f"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-win32.whl", hash = "sha256:397081c1a0bfb5124355710fe79478cdbeb39626492b15d399526ae53422b906"}, + {file = "MarkupSafe-2.1.5-cp311-cp311-win_amd64.whl", hash = "sha256:2b7c57a4dfc4f16f7142221afe5ba4e093e09e728ca65c51f5620c9aaeb9a617"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:8dec4936e9c3100156f8a2dc89c4b88d5c435175ff03413b443469c7c8c5f4d1"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:3c6b973f22eb18a789b1460b4b91bf04ae3f0c4234a0a6aa6b0a92f6f7b951d4"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ac07bad82163452a6884fe8fa0963fb98c2346ba78d779ec06bd7a6262132aee"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f5dfb42c4604dddc8e4305050aa6deb084540643ed5804d7455b5df8fe16f5e5"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ea3d8a3d18833cf4304cd2fc9cbb1efe188ca9b5efef2bdac7adc20594a0e46b"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:d050b3361367a06d752db6ead6e7edeb0009be66bc3bae0ee9d97fb326badc2a"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-musllinux_1_1_i686.whl", hash = "sha256:bec0a414d016ac1a18862a519e54b2fd0fc8bbfd6890376898a6c0891dd82e9f"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:58c98fee265677f63a4385256a6d7683ab1832f3ddd1e66fe948d5880c21a169"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-win32.whl", hash = "sha256:8590b4ae07a35970728874632fed7bd57b26b0102df2d2b233b6d9d82f6c62ad"}, + {file = "MarkupSafe-2.1.5-cp312-cp312-win_amd64.whl", hash = "sha256:823b65d8706e32ad2df51ed89496147a42a2a6e01c13cfb6ffb8b1e92bc910bb"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:c8b29db45f8fe46ad280a7294f5c3ec36dbac9491f2d1c17345be8e69cc5928f"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ec6a563cff360b50eed26f13adc43e61bc0c04d94b8be985e6fb24b81f6dcfdf"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a549b9c31bec33820e885335b451286e2969a2d9e24879f83fe904a5ce59d70a"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:4f11aa001c540f62c6166c7726f71f7573b52c68c31f014c25cc7901deea0b52"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:7b2e5a267c855eea6b4283940daa6e88a285f5f2a67f2220203786dfa59b37e9"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:2d2d793e36e230fd32babe143b04cec8a8b3eb8a3122d2aceb4a371e6b09b8df"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:ce409136744f6521e39fd8e2a24c53fa18ad67aa5bc7c2cf83645cce5b5c4e50"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-win32.whl", hash = "sha256:4096e9de5c6fdf43fb4f04c26fb114f61ef0bf2e5604b6ee3019d51b69e8c371"}, + {file = "MarkupSafe-2.1.5-cp37-cp37m-win_amd64.whl", hash = "sha256:4275d846e41ecefa46e2015117a9f491e57a71ddd59bbead77e904dc02b1bed2"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:656f7526c69fac7f600bd1f400991cc282b417d17539a1b228617081106feb4a"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:97cafb1f3cbcd3fd2b6fbfb99ae11cdb14deea0736fc2b0952ee177f2b813a46"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:1f3fbcb7ef1f16e48246f704ab79d79da8a46891e2da03f8783a5b6fa41a9532"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:fa9db3f79de01457b03d4f01b34cf91bc0048eb2c3846ff26f66687c2f6d16ab"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ffee1f21e5ef0d712f9033568f8344d5da8cc2869dbd08d87c84656e6a2d2f68"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:5dedb4db619ba5a2787a94d877bc8ffc0566f92a01c0ef214865e54ecc9ee5e0"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:30b600cf0a7ac9234b2638fbc0fb6158ba5bdcdf46aeb631ead21248b9affbc4"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:8dd717634f5a044f860435c1d8c16a270ddf0ef8588d4887037c5028b859b0c3"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-win32.whl", hash = "sha256:daa4ee5a243f0f20d528d939d06670a298dd39b1ad5f8a72a4275124a7819eff"}, + {file = "MarkupSafe-2.1.5-cp38-cp38-win_amd64.whl", hash = "sha256:619bc166c4f2de5caa5a633b8b7326fbe98e0ccbfacabd87268a2b15ff73a029"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:7a68b554d356a91cce1236aa7682dc01df0edba8d043fd1ce607c49dd3c1edcf"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:db0b55e0f3cc0be60c1f19efdde9a637c32740486004f20d1cff53c3c0ece4d2"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3e53af139f8579a6d5f7b76549125f0d94d7e630761a2111bc431fd820e163b8"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:17b950fccb810b3293638215058e432159d2b71005c74371d784862b7e4683f3"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:4c31f53cdae6ecfa91a77820e8b151dba54ab528ba65dfd235c80b086d68a465"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:bff1b4290a66b490a2f4719358c0cdcd9bafb6b8f061e45c7a2460866bf50c2e"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:bc1667f8b83f48511b94671e0e441401371dfd0f0a795c7daa4a3cd1dde55bea"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:5049256f536511ee3f7e1b3f87d1d1209d327e818e6ae1365e8653d7e3abb6a6"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-win32.whl", hash = "sha256:00e046b6dd71aa03a41079792f8473dc494d564611a8f89bbbd7cb93295ebdcf"}, + {file = "MarkupSafe-2.1.5-cp39-cp39-win_amd64.whl", hash = "sha256:fa173ec60341d6bb97a89f5ea19c85c5643c1e7dedebc22f5181eb73573142c5"}, + {file = "MarkupSafe-2.1.5.tar.gz", hash = "sha256:d283d37a890ba4c1ae73ffadf8046435c76e7bc2247bbb63c00bd1a709c6544b"}, +] + +[[package]] +name = "more-itertools" +version = "10.2.0" +description = "More routines for operating on iterables, beyond itertools" +optional = false +python-versions = ">=3.8" +files = [ + {file = "more-itertools-10.2.0.tar.gz", hash = "sha256:8fccb480c43d3e99a00087634c06dd02b0d50fbf088b380de5a41a015ec239e1"}, + {file = "more_itertools-10.2.0-py3-none-any.whl", hash = "sha256:686b06abe565edfab151cb8fd385a05651e1fdf8f0a14191e4439283421f8684"}, +] + +[[package]] +name = "mpmath" +version = "1.3.0" +description = "Python library for arbitrary-precision floating-point arithmetic" +optional = false +python-versions = "*" +files = [ + {file = "mpmath-1.3.0-py3-none-any.whl", hash = "sha256:a0b2b9fe80bbcd81a6647ff13108738cfb482d481d826cc0e02f5b35e5c88d2c"}, + {file = "mpmath-1.3.0.tar.gz", hash = "sha256:7a28eb2a9774d00c7bc92411c19a89209d5da7c4c9a9e227be8330a23a25b91f"}, +] + +[package.extras] +develop = ["codecov", "pycodestyle", "pytest (>=4.6)", "pytest-cov", "wheel"] +docs = ["sphinx"] +gmpy = ["gmpy2 (>=2.1.0a4)"] +tests = ["pytest (>=4.6)"] + +[[package]] +name = "networkx" +version = "3.2.1" +description = "Python package for creating and manipulating graphs and networks" +optional = false +python-versions = ">=3.9" +files = [ + {file = "networkx-3.2.1-py3-none-any.whl", hash = "sha256:f18c69adc97877c42332c170849c96cefa91881c99a7cb3e95b7c659ebdc1ec2"}, + {file = "networkx-3.2.1.tar.gz", hash = "sha256:9f1bb5cf3409bf324e0a722c20bdb4c20ee39bf1c30ce8ae499c8502b0b5e0c6"}, +] + +[package.extras] +default = ["matplotlib (>=3.5)", "numpy (>=1.22)", "pandas (>=1.4)", "scipy (>=1.9,!=1.11.0,!=1.11.1)"] +developer = ["changelist (==0.4)", "mypy (>=1.1)", "pre-commit (>=3.2)", "rtoml"] +doc = ["nb2plots (>=0.7)", "nbconvert (<7.9)", "numpydoc (>=1.6)", "pillow (>=9.4)", "pydata-sphinx-theme (>=0.14)", "sphinx (>=7)", "sphinx-gallery (>=0.14)", "texext (>=0.6.7)"] +extra = ["lxml (>=4.6)", "pydot (>=1.4.2)", "pygraphviz (>=1.11)", "sympy (>=1.10)"] +test = ["pytest (>=7.2)", "pytest-cov (>=4.0)"] + +[[package]] +name = "numpy" +version = "1.26.4" +description = "Fundamental package for array computing in Python" +optional = false +python-versions = ">=3.9" +files = [ + {file = "numpy-1.26.4-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:9ff0f4f29c51e2803569d7a51c2304de5554655a60c5d776e35b4a41413830d0"}, + {file = "numpy-1.26.4-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:2e4ee3380d6de9c9ec04745830fd9e2eccb3e6cf790d39d7b98ffd19b0dd754a"}, + {file = "numpy-1.26.4-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d209d8969599b27ad20994c8e41936ee0964e6da07478d6c35016bc386b66ad4"}, + {file = "numpy-1.26.4-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ffa75af20b44f8dba823498024771d5ac50620e6915abac414251bd971b4529f"}, + {file = "numpy-1.26.4-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:62b8e4b1e28009ef2846b4c7852046736bab361f7aeadeb6a5b89ebec3c7055a"}, + {file = "numpy-1.26.4-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:a4abb4f9001ad2858e7ac189089c42178fcce737e4169dc61321660f1a96c7d2"}, + {file = "numpy-1.26.4-cp310-cp310-win32.whl", hash = "sha256:bfe25acf8b437eb2a8b2d49d443800a5f18508cd811fea3181723922a8a82b07"}, + {file = "numpy-1.26.4-cp310-cp310-win_amd64.whl", hash = "sha256:b97fe8060236edf3662adfc2c633f56a08ae30560c56310562cb4f95500022d5"}, + {file = "numpy-1.26.4-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:4c66707fabe114439db9068ee468c26bbdf909cac0fb58686a42a24de1760c71"}, + {file = "numpy-1.26.4-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:edd8b5fe47dab091176d21bb6de568acdd906d1887a4584a15a9a96a1dca06ef"}, + {file = "numpy-1.26.4-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:7ab55401287bfec946ced39700c053796e7cc0e3acbef09993a9ad2adba6ca6e"}, + {file = "numpy-1.26.4-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:666dbfb6ec68962c033a450943ded891bed2d54e6755e35e5835d63f4f6931d5"}, + {file = "numpy-1.26.4-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:96ff0b2ad353d8f990b63294c8986f1ec3cb19d749234014f4e7eb0112ceba5a"}, + {file = "numpy-1.26.4-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:60dedbb91afcbfdc9bc0b1f3f402804070deed7392c23eb7a7f07fa857868e8a"}, + {file = "numpy-1.26.4-cp311-cp311-win32.whl", hash = "sha256:1af303d6b2210eb850fcf03064d364652b7120803a0b872f5211f5234b399f20"}, + {file = "numpy-1.26.4-cp311-cp311-win_amd64.whl", hash = "sha256:cd25bcecc4974d09257ffcd1f098ee778f7834c3ad767fe5db785be9a4aa9cb2"}, + {file = "numpy-1.26.4-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:b3ce300f3644fb06443ee2222c2201dd3a89ea6040541412b8fa189341847218"}, + {file = "numpy-1.26.4-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:03a8c78d01d9781b28a6989f6fa1bb2c4f2d51201cf99d3dd875df6fbd96b23b"}, + {file = "numpy-1.26.4-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:9fad7dcb1aac3c7f0584a5a8133e3a43eeb2fe127f47e3632d43d677c66c102b"}, + {file = "numpy-1.26.4-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:675d61ffbfa78604709862923189bad94014bef562cc35cf61d3a07bba02a7ed"}, + {file = "numpy-1.26.4-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:ab47dbe5cc8210f55aa58e4805fe224dac469cde56b9f731a4c098b91917159a"}, + {file = "numpy-1.26.4-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:1dda2e7b4ec9dd512f84935c5f126c8bd8b9f2fc001e9f54af255e8c5f16b0e0"}, + {file = "numpy-1.26.4-cp312-cp312-win32.whl", hash = "sha256:50193e430acfc1346175fcbdaa28ffec49947a06918b7b92130744e81e640110"}, + {file = "numpy-1.26.4-cp312-cp312-win_amd64.whl", hash = "sha256:08beddf13648eb95f8d867350f6a018a4be2e5ad54c8d8caed89ebca558b2818"}, + {file = "numpy-1.26.4-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:7349ab0fa0c429c82442a27a9673fc802ffdb7c7775fad780226cb234965e53c"}, + {file = "numpy-1.26.4-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:52b8b60467cd7dd1e9ed082188b4e6bb35aa5cdd01777621a1658910745b90be"}, + {file = "numpy-1.26.4-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d5241e0a80d808d70546c697135da2c613f30e28251ff8307eb72ba696945764"}, + {file = "numpy-1.26.4-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f870204a840a60da0b12273ef34f7051e98c3b5961b61b0c2c1be6dfd64fbcd3"}, + {file = "numpy-1.26.4-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:679b0076f67ecc0138fd2ede3a8fd196dddc2ad3254069bcb9faf9a79b1cebcd"}, + {file = "numpy-1.26.4-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:47711010ad8555514b434df65f7d7b076bb8261df1ca9bb78f53d3b2db02e95c"}, + {file = "numpy-1.26.4-cp39-cp39-win32.whl", hash = "sha256:a354325ee03388678242a4d7ebcd08b5c727033fcff3b2f536aea978e15ee9e6"}, + {file = "numpy-1.26.4-cp39-cp39-win_amd64.whl", hash = "sha256:3373d5d70a5fe74a2c1bb6d2cfd9609ecf686d47a2d7b1d37a8f3b6bf6003aea"}, + {file = "numpy-1.26.4-pp39-pypy39_pp73-macosx_10_9_x86_64.whl", hash = "sha256:afedb719a9dcfc7eaf2287b839d8198e06dcd4cb5d276a3df279231138e83d30"}, + {file = "numpy-1.26.4-pp39-pypy39_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:95a7476c59002f2f6c590b9b7b998306fba6a5aa646b1e22ddfeaf8f78c3a29c"}, + {file = "numpy-1.26.4-pp39-pypy39_pp73-win_amd64.whl", hash = "sha256:7e50d0a0cc3189f9cb0aeb3a6a6af18c16f59f004b866cd2be1c14b36134a4a0"}, + {file = "numpy-1.26.4.tar.gz", hash = "sha256:2a02aba9ed12e4ac4eb3ea9421c420301a0c6460d9830d74a9df87efa4912010"}, +] + +[[package]] +name = "packaging" +version = "23.2" +description = "Core utilities for Python packages" +optional = false +python-versions = ">=3.7" +files = [ + {file = "packaging-23.2-py3-none-any.whl", hash = "sha256:8c491190033a9af7e1d931d0b5dacc2ef47509b34dd0de67ed209b5203fc88c7"}, + {file = "packaging-23.2.tar.gz", hash = "sha256:048fb0e9405036518eaaf48a55953c750c11e1a1b68e0dd1a9d62ed0c092cfc5"}, +] + +[[package]] +name = "pluggy" +version = "0.13.1" +description = "plugin and hook calling mechanisms for python" +optional = false +python-versions = ">=2.7, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*" +files = [ + {file = "pluggy-0.13.1-py2.py3-none-any.whl", hash = "sha256:966c145cd83c96502c3c3868f50408687b38434af77734af1e9ca461a4081d2d"}, + {file = "pluggy-0.13.1.tar.gz", hash = "sha256:15b2acde666561e1298d71b523007ed7364de07029219b604cf808bfa1c765b0"}, +] + +[package.extras] +dev = ["pre-commit", "tox"] + +[[package]] +name = "protobuf" +version = "4.25.3" +description = "" +optional = false +python-versions = ">=3.8" +files = [ + {file = "protobuf-4.25.3-cp310-abi3-win32.whl", hash = "sha256:d4198877797a83cbfe9bffa3803602bbe1625dc30d8a097365dbc762e5790faa"}, + {file = "protobuf-4.25.3-cp310-abi3-win_amd64.whl", hash = "sha256:209ba4cc916bab46f64e56b85b090607a676f66b473e6b762e6f1d9d591eb2e8"}, + {file = "protobuf-4.25.3-cp37-abi3-macosx_10_9_universal2.whl", hash = "sha256:f1279ab38ecbfae7e456a108c5c0681e4956d5b1090027c1de0f934dfdb4b35c"}, + {file = "protobuf-4.25.3-cp37-abi3-manylinux2014_aarch64.whl", hash = "sha256:e7cb0ae90dd83727f0c0718634ed56837bfeeee29a5f82a7514c03ee1364c019"}, + {file = "protobuf-4.25.3-cp37-abi3-manylinux2014_x86_64.whl", hash = "sha256:7c8daa26095f82482307bc717364e7c13f4f1c99659be82890dcfc215194554d"}, + {file = "protobuf-4.25.3-cp38-cp38-win32.whl", hash = "sha256:f4f118245c4a087776e0a8408be33cf09f6c547442c00395fbfb116fac2f8ac2"}, + {file = "protobuf-4.25.3-cp38-cp38-win_amd64.whl", hash = "sha256:c053062984e61144385022e53678fbded7aea14ebb3e0305ae3592fb219ccfa4"}, + {file = "protobuf-4.25.3-cp39-cp39-win32.whl", hash = "sha256:19b270aeaa0099f16d3ca02628546b8baefe2955bbe23224aaf856134eccf1e4"}, + {file = "protobuf-4.25.3-cp39-cp39-win_amd64.whl", hash = "sha256:e3c97a1555fd6388f857770ff8b9703083de6bf1f9274a002a332d65fbb56c8c"}, + {file = "protobuf-4.25.3-py3-none-any.whl", hash = "sha256:f0700d54bcf45424477e46a9f0944155b46fb0639d69728739c0e47bab83f2b9"}, + {file = "protobuf-4.25.3.tar.gz", hash = "sha256:25b5d0b42fd000320bd7830b349e3b696435f3b329810427a6bcce6a5492cc5c"}, +] + +[[package]] +name = "py" +version = "1.11.0" +description = "library with cross-python path, ini-parsing, io, code, log facilities" +optional = false +python-versions = ">=2.7, !=3.0.*, !=3.1.*, !=3.2.*, !=3.3.*, !=3.4.*" +files = [ + {file = "py-1.11.0-py2.py3-none-any.whl", hash = "sha256:607c53218732647dff4acdfcd50cb62615cedf612e72d1724fb1a0cc6405b378"}, + {file = "py-1.11.0.tar.gz", hash = "sha256:51c75c4126074b472f746a24399ad32f6053d1b34b68d2fa41e558e6f4a98719"}, +] + +[[package]] +name = "pytest" +version = "5.4.3" +description = "pytest: simple powerful testing with Python" +optional = false +python-versions = ">=3.5" +files = [ + {file = "pytest-5.4.3-py3-none-any.whl", hash = "sha256:5c0db86b698e8f170ba4582a492248919255fcd4c79b1ee64ace34301fb589a1"}, + {file = "pytest-5.4.3.tar.gz", hash = "sha256:7979331bfcba207414f5e1263b5a0f8f521d0f457318836a7355531ed1a4c7d8"}, +] + +[package.dependencies] +atomicwrites = {version = ">=1.0", markers = "sys_platform == \"win32\""} +attrs = ">=17.4.0" +colorama = {version = "*", markers = "sys_platform == \"win32\""} +more-itertools = ">=4.0.0" +packaging = "*" +pluggy = ">=0.12,<1.0" +py = ">=1.5.0" +wcwidth = "*" + +[package.extras] +checkqa-mypy = ["mypy (==v0.761)"] +testing = ["argcomplete", "hypothesis (>=3.56)", "mock", "nose", "requests", "xmlschema"] + +[[package]] +name = "pyyaml" +version = "6.0.1" +description = "YAML parser and emitter for Python" +optional = false +python-versions = ">=3.6" +files = [ + {file = "PyYAML-6.0.1-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:d858aa552c999bc8a8d57426ed01e40bef403cd8ccdd0fc5f6f04a00414cac2a"}, + {file = "PyYAML-6.0.1-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:fd66fc5d0da6d9815ba2cebeb4205f95818ff4b79c3ebe268e75d961704af52f"}, + {file = "PyYAML-6.0.1-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:69b023b2b4daa7548bcfbd4aa3da05b3a74b772db9e23b982788168117739938"}, + {file = "PyYAML-6.0.1-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:81e0b275a9ecc9c0c0c07b4b90ba548307583c125f54d5b6946cfee6360c733d"}, + {file = "PyYAML-6.0.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ba336e390cd8e4d1739f42dfe9bb83a3cc2e80f567d8805e11b46f4a943f5515"}, + {file = "PyYAML-6.0.1-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:326c013efe8048858a6d312ddd31d56e468118ad4cdeda36c719bf5bb6192290"}, + {file = "PyYAML-6.0.1-cp310-cp310-win32.whl", hash = "sha256:bd4af7373a854424dabd882decdc5579653d7868b8fb26dc7d0e99f823aa5924"}, + {file = "PyYAML-6.0.1-cp310-cp310-win_amd64.whl", hash = "sha256:fd1592b3fdf65fff2ad0004b5e363300ef59ced41c2e6b3a99d4089fa8c5435d"}, + {file = "PyYAML-6.0.1-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:6965a7bc3cf88e5a1c3bd2e0b5c22f8d677dc88a455344035f03399034eb3007"}, + {file = "PyYAML-6.0.1-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:f003ed9ad21d6a4713f0a9b5a7a0a79e08dd0f221aff4525a2be4c346ee60aab"}, + {file = "PyYAML-6.0.1-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:42f8152b8dbc4fe7d96729ec2b99c7097d656dc1213a3229ca5383f973a5ed6d"}, + {file = "PyYAML-6.0.1-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:062582fca9fabdd2c8b54a3ef1c978d786e0f6b3a1510e0ac93ef59e0ddae2bc"}, + {file = "PyYAML-6.0.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:d2b04aac4d386b172d5b9692e2d2da8de7bfb6c387fa4f801fbf6fb2e6ba4673"}, + {file = "PyYAML-6.0.1-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:e7d73685e87afe9f3b36c799222440d6cf362062f78be1013661b00c5c6f678b"}, + {file = "PyYAML-6.0.1-cp311-cp311-win32.whl", hash = "sha256:1635fd110e8d85d55237ab316b5b011de701ea0f29d07611174a1b42f1444741"}, + {file = "PyYAML-6.0.1-cp311-cp311-win_amd64.whl", hash = "sha256:bf07ee2fef7014951eeb99f56f39c9bb4af143d8aa3c21b1677805985307da34"}, + {file = "PyYAML-6.0.1-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:855fb52b0dc35af121542a76b9a84f8d1cd886ea97c84703eaa6d88e37a2ad28"}, + {file = "PyYAML-6.0.1-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:40df9b996c2b73138957fe23a16a4f0ba614f4c0efce1e9406a184b6d07fa3a9"}, + {file = "PyYAML-6.0.1-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a08c6f0fe150303c1c6b71ebcd7213c2858041a7e01975da3a99aed1e7a378ef"}, + {file = "PyYAML-6.0.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:6c22bec3fbe2524cde73d7ada88f6566758a8f7227bfbf93a408a9d86bcc12a0"}, + {file = "PyYAML-6.0.1-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:8d4e9c88387b0f5c7d5f281e55304de64cf7f9c0021a3525bd3b1c542da3b0e4"}, + {file = "PyYAML-6.0.1-cp312-cp312-win32.whl", hash = "sha256:d483d2cdf104e7c9fa60c544d92981f12ad66a457afae824d146093b8c294c54"}, + {file = "PyYAML-6.0.1-cp312-cp312-win_amd64.whl", hash = "sha256:0d3304d8c0adc42be59c5f8a4d9e3d7379e6955ad754aa9d6ab7a398b59dd1df"}, + {file = "PyYAML-6.0.1-cp36-cp36m-macosx_10_9_x86_64.whl", hash = "sha256:50550eb667afee136e9a77d6dc71ae76a44df8b3e51e41b77f6de2932bfe0f47"}, + {file = "PyYAML-6.0.1-cp36-cp36m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:1fe35611261b29bd1de0070f0b2f47cb6ff71fa6595c077e42bd0c419fa27b98"}, + {file = "PyYAML-6.0.1-cp36-cp36m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:704219a11b772aea0d8ecd7058d0082713c3562b4e271b849ad7dc4a5c90c13c"}, + {file = "PyYAML-6.0.1-cp36-cp36m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:afd7e57eddb1a54f0f1a974bc4391af8bcce0b444685d936840f125cf046d5bd"}, + {file = "PyYAML-6.0.1-cp36-cp36m-win32.whl", hash = "sha256:fca0e3a251908a499833aa292323f32437106001d436eca0e6e7833256674585"}, + {file = "PyYAML-6.0.1-cp36-cp36m-win_amd64.whl", hash = "sha256:f22ac1c3cac4dbc50079e965eba2c1058622631e526bd9afd45fedd49ba781fa"}, + {file = "PyYAML-6.0.1-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:b1275ad35a5d18c62a7220633c913e1b42d44b46ee12554e5fd39c70a243d6a3"}, + {file = "PyYAML-6.0.1-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:18aeb1bf9a78867dc38b259769503436b7c72f7a1f1f4c93ff9a17de54319b27"}, + {file = "PyYAML-6.0.1-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:596106435fa6ad000c2991a98fa58eeb8656ef2325d7e158344fb33864ed87e3"}, + {file = "PyYAML-6.0.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:baa90d3f661d43131ca170712d903e6295d1f7a0f595074f151c0aed377c9b9c"}, + {file = "PyYAML-6.0.1-cp37-cp37m-win32.whl", hash = "sha256:9046c58c4395dff28dd494285c82ba00b546adfc7ef001486fbf0324bc174fba"}, + {file = "PyYAML-6.0.1-cp37-cp37m-win_amd64.whl", hash = "sha256:4fb147e7a67ef577a588a0e2c17b6db51dda102c71de36f8549b6816a96e1867"}, + {file = "PyYAML-6.0.1-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:1d4c7e777c441b20e32f52bd377e0c409713e8bb1386e1099c2415f26e479595"}, + {file = "PyYAML-6.0.1-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a0cd17c15d3bb3fa06978b4e8958dcdc6e0174ccea823003a106c7d4d7899ac5"}, + {file = "PyYAML-6.0.1-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:28c119d996beec18c05208a8bd78cbe4007878c6dd15091efb73a30e90539696"}, + {file = "PyYAML-6.0.1-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:7e07cbde391ba96ab58e532ff4803f79c4129397514e1413a7dc761ccd755735"}, + {file = "PyYAML-6.0.1-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:49a183be227561de579b4a36efbb21b3eab9651dd81b1858589f796549873dd6"}, + {file = "PyYAML-6.0.1-cp38-cp38-win32.whl", hash = "sha256:184c5108a2aca3c5b3d3bf9395d50893a7ab82a38004c8f61c258d4428e80206"}, + {file = "PyYAML-6.0.1-cp38-cp38-win_amd64.whl", hash = "sha256:1e2722cc9fbb45d9b87631ac70924c11d3a401b2d7f410cc0e3bbf249f2dca62"}, + {file = "PyYAML-6.0.1-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:9eb6caa9a297fc2c2fb8862bc5370d0303ddba53ba97e71f08023b6cd73d16a8"}, + {file = "PyYAML-6.0.1-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:c8098ddcc2a85b61647b2590f825f3db38891662cfc2fc776415143f599bb859"}, + {file = "PyYAML-6.0.1-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:5773183b6446b2c99bb77e77595dd486303b4faab2b086e7b17bc6bef28865f6"}, + {file = "PyYAML-6.0.1-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:b786eecbdf8499b9ca1d697215862083bd6d2a99965554781d0d8d1ad31e13a0"}, + {file = "PyYAML-6.0.1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bc1bf2925a1ecd43da378f4db9e4f799775d6367bdb94671027b73b393a7c42c"}, + {file = "PyYAML-6.0.1-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:04ac92ad1925b2cff1db0cfebffb6ffc43457495c9b3c39d3fcae417d7125dc5"}, + {file = "PyYAML-6.0.1-cp39-cp39-win32.whl", hash = "sha256:faca3bdcf85b2fc05d06ff3fbc1f83e1391b3e724afa3feba7d13eeab355484c"}, + {file = "PyYAML-6.0.1-cp39-cp39-win_amd64.whl", hash = "sha256:510c9deebc5c0225e8c96813043e62b680ba2f9c50a08d3724c7f28a747d1486"}, + {file = "PyYAML-6.0.1.tar.gz", hash = "sha256:bfdf460b1736c775f2ba9f6a92bca30bc2095067b8a9d77876d1fad6cc3b4a43"}, +] + +[[package]] +name = "regex" +version = "2023.12.25" +description = "Alternative regular expression module, to replace re." +optional = false +python-versions = ">=3.7" +files = [ + {file = "regex-2023.12.25-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:0694219a1d54336fd0445ea382d49d36882415c0134ee1e8332afd1529f0baa5"}, + {file = "regex-2023.12.25-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:b014333bd0217ad3d54c143de9d4b9a3ca1c5a29a6d0d554952ea071cff0f1f8"}, + {file = "regex-2023.12.25-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:d865984b3f71f6d0af64d0d88f5733521698f6c16f445bb09ce746c92c97c586"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:1e0eabac536b4cc7f57a5f3d095bfa557860ab912f25965e08fe1545e2ed8b4c"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:c25a8ad70e716f96e13a637802813f65d8a6760ef48672aa3502f4c24ea8b400"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a9b6d73353f777630626f403b0652055ebfe8ff142a44ec2cf18ae470395766e"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a9cc99d6946d750eb75827cb53c4371b8b0fe89c733a94b1573c9dd16ea6c9e4"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:88d1f7bef20c721359d8675f7d9f8e414ec5003d8f642fdfd8087777ff7f94b5"}, + {file = "regex-2023.12.25-cp310-cp310-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:cb3fe77aec8f1995611f966d0c656fdce398317f850d0e6e7aebdfe61f40e1cd"}, + {file = "regex-2023.12.25-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:7aa47c2e9ea33a4a2a05f40fcd3ea36d73853a2aae7b4feab6fc85f8bf2c9704"}, + {file = "regex-2023.12.25-cp310-cp310-musllinux_1_1_i686.whl", hash = "sha256:df26481f0c7a3f8739fecb3e81bc9da3fcfae34d6c094563b9d4670b047312e1"}, + {file = "regex-2023.12.25-cp310-cp310-musllinux_1_1_ppc64le.whl", hash = "sha256:c40281f7d70baf6e0db0c2f7472b31609f5bc2748fe7275ea65a0b4601d9b392"}, + {file = "regex-2023.12.25-cp310-cp310-musllinux_1_1_s390x.whl", hash = "sha256:d94a1db462d5690ebf6ae86d11c5e420042b9898af5dcf278bd97d6bda065423"}, + {file = "regex-2023.12.25-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:ba1b30765a55acf15dce3f364e4928b80858fa8f979ad41f862358939bdd1f2f"}, + {file = "regex-2023.12.25-cp310-cp310-win32.whl", hash = "sha256:150c39f5b964e4d7dba46a7962a088fbc91f06e606f023ce57bb347a3b2d4630"}, + {file = "regex-2023.12.25-cp310-cp310-win_amd64.whl", hash = "sha256:09da66917262d9481c719599116c7dc0c321ffcec4b1f510c4f8a066f8768105"}, + {file = "regex-2023.12.25-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:1b9d811f72210fa9306aeb88385b8f8bcef0dfbf3873410413c00aa94c56c2b6"}, + {file = "regex-2023.12.25-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:d902a43085a308cef32c0d3aea962524b725403fd9373dea18110904003bac97"}, + {file = "regex-2023.12.25-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:d166eafc19f4718df38887b2bbe1467a4f74a9830e8605089ea7a30dd4da8887"}, + {file = "regex-2023.12.25-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c7ad32824b7f02bb3c9f80306d405a1d9b7bb89362d68b3c5a9be53836caebdb"}, + {file = "regex-2023.12.25-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:636ba0a77de609d6510235b7f0e77ec494d2657108f777e8765efc060094c98c"}, + {file = "regex-2023.12.25-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:0fda75704357805eb953a3ee15a2b240694a9a514548cd49b3c5124b4e2ad01b"}, + {file = "regex-2023.12.25-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f72cbae7f6b01591f90814250e636065850c5926751af02bb48da94dfced7baa"}, + {file = "regex-2023.12.25-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:db2a0b1857f18b11e3b0e54ddfefc96af46b0896fb678c85f63fb8c37518b3e7"}, + {file = "regex-2023.12.25-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:7502534e55c7c36c0978c91ba6f61703faf7ce733715ca48f499d3dbbd7657e0"}, + {file = "regex-2023.12.25-cp311-cp311-musllinux_1_1_i686.whl", hash = "sha256:e8c7e08bb566de4faaf11984af13f6bcf6a08f327b13631d41d62592681d24fe"}, + {file = "regex-2023.12.25-cp311-cp311-musllinux_1_1_ppc64le.whl", hash = "sha256:283fc8eed679758de38fe493b7d7d84a198b558942b03f017b1f94dda8efae80"}, + {file = "regex-2023.12.25-cp311-cp311-musllinux_1_1_s390x.whl", hash = "sha256:f44dd4d68697559d007462b0a3a1d9acd61d97072b71f6d1968daef26bc744bd"}, + {file = "regex-2023.12.25-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:67d3ccfc590e5e7197750fcb3a2915b416a53e2de847a728cfa60141054123d4"}, + {file = "regex-2023.12.25-cp311-cp311-win32.whl", hash = "sha256:68191f80a9bad283432385961d9efe09d783bcd36ed35a60fb1ff3f1ec2efe87"}, + {file = "regex-2023.12.25-cp311-cp311-win_amd64.whl", hash = "sha256:7d2af3f6b8419661a0c421584cfe8aaec1c0e435ce7e47ee2a97e344b98f794f"}, + {file = "regex-2023.12.25-cp312-cp312-macosx_10_9_universal2.whl", hash = "sha256:8a0ccf52bb37d1a700375a6b395bff5dd15c50acb745f7db30415bae3c2b0715"}, + {file = "regex-2023.12.25-cp312-cp312-macosx_10_9_x86_64.whl", hash = "sha256:c3c4a78615b7762740531c27cf46e2f388d8d727d0c0c739e72048beb26c8a9d"}, + {file = "regex-2023.12.25-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:ad83e7545b4ab69216cef4cc47e344d19622e28aabec61574b20257c65466d6a"}, + {file = "regex-2023.12.25-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:b7a635871143661feccce3979e1727c4e094f2bdfd3ec4b90dfd4f16f571a87a"}, + {file = "regex-2023.12.25-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:d498eea3f581fbe1b34b59c697512a8baef88212f92e4c7830fcc1499f5b45a5"}, + {file = "regex-2023.12.25-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:43f7cd5754d02a56ae4ebb91b33461dc67be8e3e0153f593c509e21d219c5060"}, + {file = "regex-2023.12.25-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:51f4b32f793812714fd5307222a7f77e739b9bc566dc94a18126aba3b92b98a3"}, + {file = "regex-2023.12.25-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ba99d8077424501b9616b43a2d208095746fb1284fc5ba490139651f971d39d9"}, + {file = "regex-2023.12.25-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:4bfc2b16e3ba8850e0e262467275dd4d62f0d045e0e9eda2bc65078c0110a11f"}, + {file = "regex-2023.12.25-cp312-cp312-musllinux_1_1_i686.whl", hash = "sha256:8c2c19dae8a3eb0ea45a8448356ed561be843b13cbc34b840922ddf565498c1c"}, + {file = "regex-2023.12.25-cp312-cp312-musllinux_1_1_ppc64le.whl", hash = "sha256:60080bb3d8617d96f0fb7e19796384cc2467447ef1c491694850ebd3670bc457"}, + {file = "regex-2023.12.25-cp312-cp312-musllinux_1_1_s390x.whl", hash = "sha256:b77e27b79448e34c2c51c09836033056a0547aa360c45eeeb67803da7b0eedaf"}, + {file = "regex-2023.12.25-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:518440c991f514331f4850a63560321f833979d145d7d81186dbe2f19e27ae3d"}, + {file = "regex-2023.12.25-cp312-cp312-win32.whl", hash = "sha256:e2610e9406d3b0073636a3a2e80db05a02f0c3169b5632022b4e81c0364bcda5"}, + {file = "regex-2023.12.25-cp312-cp312-win_amd64.whl", hash = "sha256:cc37b9aeebab425f11f27e5e9e6cf580be7206c6582a64467a14dda211abc232"}, + {file = "regex-2023.12.25-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:da695d75ac97cb1cd725adac136d25ca687da4536154cdc2815f576e4da11c69"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d126361607b33c4eb7b36debc173bf25d7805847346dd4d99b5499e1fef52bc7"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:4719bb05094d7d8563a450cf8738d2e1061420f79cfcc1fa7f0a44744c4d8f73"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:5dd58946bce44b53b06d94aa95560d0b243eb2fe64227cba50017a8d8b3cd3e2"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:22a86d9fff2009302c440b9d799ef2fe322416d2d58fc124b926aa89365ec482"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:2aae8101919e8aa05ecfe6322b278f41ce2994c4a430303c4cd163fef746e04f"}, + {file = "regex-2023.12.25-cp37-cp37m-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:e692296c4cc2873967771345a876bcfc1c547e8dd695c6b89342488b0ea55cd8"}, + {file = "regex-2023.12.25-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:263ef5cc10979837f243950637fffb06e8daed7f1ac1e39d5910fd29929e489a"}, + {file = "regex-2023.12.25-cp37-cp37m-musllinux_1_1_i686.whl", hash = "sha256:d6f7e255e5fa94642a0724e35406e6cb7001c09d476ab5fce002f652b36d0c39"}, + {file = "regex-2023.12.25-cp37-cp37m-musllinux_1_1_ppc64le.whl", hash = "sha256:88ad44e220e22b63b0f8f81f007e8abbb92874d8ced66f32571ef8beb0643b2b"}, + {file = "regex-2023.12.25-cp37-cp37m-musllinux_1_1_s390x.whl", hash = "sha256:3a17d3ede18f9cedcbe23d2daa8a2cd6f59fe2bf082c567e43083bba3fb00347"}, + {file = "regex-2023.12.25-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:d15b274f9e15b1a0b7a45d2ac86d1f634d983ca40d6b886721626c47a400bf39"}, + {file = "regex-2023.12.25-cp37-cp37m-win32.whl", hash = "sha256:ed19b3a05ae0c97dd8f75a5d8f21f7723a8c33bbc555da6bbe1f96c470139d3c"}, + {file = "regex-2023.12.25-cp37-cp37m-win_amd64.whl", hash = "sha256:a6d1047952c0b8104a1d371f88f4ab62e6275567d4458c1e26e9627ad489b445"}, + {file = "regex-2023.12.25-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:b43523d7bc2abd757119dbfb38af91b5735eea45537ec6ec3a5ec3f9562a1c53"}, + {file = "regex-2023.12.25-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:efb2d82f33b2212898f1659fb1c2e9ac30493ac41e4d53123da374c3b5541e64"}, + {file = "regex-2023.12.25-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:b7fca9205b59c1a3d5031f7e64ed627a1074730a51c2a80e97653e3e9fa0d415"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:086dd15e9435b393ae06f96ab69ab2d333f5d65cbe65ca5a3ef0ec9564dfe770"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:e81469f7d01efed9b53740aedd26085f20d49da65f9c1f41e822a33992cb1590"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:34e4af5b27232f68042aa40a91c3b9bb4da0eeb31b7632e0091afc4310afe6cb"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9852b76ab558e45b20bf1893b59af64a28bd3820b0c2efc80e0a70a4a3ea51c1"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:ff100b203092af77d1a5a7abe085b3506b7eaaf9abf65b73b7d6905b6cb76988"}, + {file = "regex-2023.12.25-cp38-cp38-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:cc038b2d8b1470364b1888a98fd22d616fba2b6309c5b5f181ad4483e0017861"}, + {file = "regex-2023.12.25-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:094ba386bb5c01e54e14434d4caabf6583334090865b23ef58e0424a6286d3dc"}, + {file = "regex-2023.12.25-cp38-cp38-musllinux_1_1_i686.whl", hash = "sha256:5cd05d0f57846d8ba4b71d9c00f6f37d6b97d5e5ef8b3c3840426a475c8f70f4"}, + {file = "regex-2023.12.25-cp38-cp38-musllinux_1_1_ppc64le.whl", hash = "sha256:9aa1a67bbf0f957bbe096375887b2505f5d8ae16bf04488e8b0f334c36e31360"}, + {file = "regex-2023.12.25-cp38-cp38-musllinux_1_1_s390x.whl", hash = "sha256:98a2636994f943b871786c9e82bfe7883ecdaba2ef5df54e1450fa9869d1f756"}, + {file = "regex-2023.12.25-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:37f8e93a81fc5e5bd8db7e10e62dc64261bcd88f8d7e6640aaebe9bc180d9ce2"}, + {file = "regex-2023.12.25-cp38-cp38-win32.whl", hash = "sha256:d78bd484930c1da2b9679290a41cdb25cc127d783768a0369d6b449e72f88beb"}, + {file = "regex-2023.12.25-cp38-cp38-win_amd64.whl", hash = "sha256:b521dcecebc5b978b447f0f69b5b7f3840eac454862270406a39837ffae4e697"}, + {file = "regex-2023.12.25-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:f7bc09bc9c29ebead055bcba136a67378f03d66bf359e87d0f7c759d6d4ffa31"}, + {file = "regex-2023.12.25-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:e14b73607d6231f3cc4622809c196b540a6a44e903bcfad940779c80dffa7be7"}, + {file = "regex-2023.12.25-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:9eda5f7a50141291beda3edd00abc2d4a5b16c29c92daf8d5bd76934150f3edc"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:cc6bb9aa69aacf0f6032c307da718f61a40cf970849e471254e0e91c56ffca95"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:298dc6354d414bc921581be85695d18912bea163a8b23cac9a2562bbcd5088b1"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:2f4e475a80ecbd15896a976aa0b386c5525d0ed34d5c600b6d3ebac0a67c7ddf"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:531ac6cf22b53e0696f8e1d56ce2396311254eb806111ddd3922c9d937151dae"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:22f3470f7524b6da61e2020672df2f3063676aff444db1daa283c2ea4ed259d6"}, + {file = "regex-2023.12.25-cp39-cp39-manylinux_2_5_x86_64.manylinux1_x86_64.manylinux_2_12_x86_64.manylinux2010_x86_64.whl", hash = "sha256:89723d2112697feaa320c9d351e5f5e7b841e83f8b143dba8e2d2b5f04e10923"}, + {file = "regex-2023.12.25-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:0ecf44ddf9171cd7566ef1768047f6e66975788258b1c6c6ca78098b95cf9a3d"}, + {file = "regex-2023.12.25-cp39-cp39-musllinux_1_1_i686.whl", hash = "sha256:905466ad1702ed4acfd67a902af50b8db1feeb9781436372261808df7a2a7bca"}, + {file = "regex-2023.12.25-cp39-cp39-musllinux_1_1_ppc64le.whl", hash = "sha256:4558410b7a5607a645e9804a3e9dd509af12fb72b9825b13791a37cd417d73a5"}, + {file = "regex-2023.12.25-cp39-cp39-musllinux_1_1_s390x.whl", hash = "sha256:7e316026cc1095f2a3e8cc012822c99f413b702eaa2ca5408a513609488cb62f"}, + {file = "regex-2023.12.25-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:3b1de218d5375cd6ac4b5493e0b9f3df2be331e86520f23382f216c137913d20"}, + {file = "regex-2023.12.25-cp39-cp39-win32.whl", hash = "sha256:11a963f8e25ab5c61348d090bf1b07f1953929c13bd2309a0662e9ff680763c9"}, + {file = "regex-2023.12.25-cp39-cp39-win_amd64.whl", hash = "sha256:e693e233ac92ba83a87024e1d32b5f9ab15ca55ddd916d878146f4e3406b5c91"}, + {file = "regex-2023.12.25.tar.gz", hash = "sha256:29171aa128da69afdf4bde412d5bedc335f2ca8fcfe4489038577d05f16181e5"}, +] + +[[package]] +name = "requests" +version = "2.31.0" +description = "Python HTTP for Humans." +optional = false +python-versions = ">=3.7" +files = [ + {file = "requests-2.31.0-py3-none-any.whl", hash = "sha256:58cd2187c01e70e6e26505bca751777aa9f2ee0b7f4300988b709f44e013003f"}, + {file = "requests-2.31.0.tar.gz", hash = "sha256:942c5a758f98d790eaed1a29cb6eefc7ffb0d1cf7af05c3d2791656dbd6ad1e1"}, +] + +[package.dependencies] +certifi = ">=2017.4.17" +charset-normalizer = ">=2,<4" +idna = ">=2.5,<4" +urllib3 = ">=1.21.1,<3" + +[package.extras] +socks = ["PySocks (>=1.5.6,!=1.5.7)"] +use-chardet-on-py3 = ["chardet (>=3.0.2,<6)"] + +[[package]] +name = "safetensors" +version = "0.4.2" +description = "" +optional = false +python-versions = ">=3.7" +files = [ + {file = "safetensors-0.4.2-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:69d8bb8384dc2cb5b72c36c4d6980771b293d1a1377b378763f5e37b6bb8d133"}, + {file = "safetensors-0.4.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:3d420e19fcef96d0067f4de4699682b4bbd85fc8fea0bd45fcd961fdf3e8c82c"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:9ca54742122fa3c4821754adb67318e1cd25c3a22bbf0c5520d5176e77a099ac"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:8b47aa643afdfd66cf7ce4c184092ae734e15d10aba2c2948f24270211801c3c"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:d88a16bbc330f27e7f2d4caaf6fb061ad0b8a756ecc4033260b0378e128ce8a2"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:e9223b8ac21085db614a510eb3445e7083cae915a9202357555fa939695d4f57"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ce6cb86133dc8930a7ab5e7438545a7f205f7a1cdd5aaf108c1d0da6bdcfbc2b"}, + {file = "safetensors-0.4.2-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:b8a628e0ae2bbc334b62952c384aa5f41621d01850f8d67b04a96b9c39dd7326"}, + {file = "safetensors-0.4.2-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:88d6beb7f811a081e0e5f1d9669fdac816c45340c04b1eaf7ebfda0ce93ea403"}, + {file = "safetensors-0.4.2-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:b57fc5b1b54cb12d8690a58a4cf4b7144730d4bde9d98aa0e1dab6295a1cd579"}, + {file = "safetensors-0.4.2-cp310-none-win32.whl", hash = "sha256:9d87a1c98803c16cf113b9ba03f07b2dce5e8eabfd1811a7f7323fcaa2a1bf47"}, + {file = "safetensors-0.4.2-cp310-none-win_amd64.whl", hash = "sha256:18930ec1d1ecb526d3d9835abc2489b8f1530877518f0c541e77ef0b7abcbd99"}, + {file = "safetensors-0.4.2-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:c5dd2ed788730ed56b415d1a11c62026b8cc8c573f55a2092afb3ab383e94fff"}, + {file = "safetensors-0.4.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:cc41791b33efb9c83a59b731619f3d15f543dfe71f3a793cb8fbf9bd5d0d5d71"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4c888bf71d5ca12a720f1ed87d407c4918afa022fb247a6546d8fac15b1f112b"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:e6b2feb4b47226a16a792e6fac3f49442714884a3d4c1008569d5068a3941be9"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f41cc0ee4b838ae8f4d8364a1b162067693d11a3893f0863be8c228d40e4d0ee"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:51b7228e46c0a483c40ba4b9470dea00fb1ff8685026bb4766799000f6328ac2"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:02697f8f2be8ca3c37a4958702dbdb1864447ef765e18b5328a1617022dcf164"}, + {file = "safetensors-0.4.2-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:27fd8f65cf7c80e4280cae1ee6bcd85c483882f6580821abe71ee1a0d3dcfca7"}, + {file = "safetensors-0.4.2-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:c487b5f113b0924c9534a07dc034830fb4ef05ce9bb6d78cfe016a7dedfe281f"}, + {file = "safetensors-0.4.2-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:da7f6483f3fe67ff39b3a55552552c67930ea10a36e9f2539d36fc205273d767"}, + {file = "safetensors-0.4.2-cp311-none-win32.whl", hash = "sha256:52a7012f6cb9cb4a132760b6308daede18a9f5f8952ce08adc7c67a7d865c2d8"}, + {file = "safetensors-0.4.2-cp311-none-win_amd64.whl", hash = "sha256:4d1361a097ac430b310ce9eed8ed4746edee33ddafdfbb965debc8966fc34dc2"}, + {file = "safetensors-0.4.2-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:77af8aa0edcc2863760fd6febbfdb82e88fd75d0e60c1ce4ba57208ba5e4a89b"}, + {file = "safetensors-0.4.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:846666c1c5a8c8888d2dfda8d3921cb9cb8e2c5f78365be756c11021e75a0a2a"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:4f4bfc7ea19b446bfad41510d4b4c76101698c00caaa8a332c8edd8090a412ef"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:233436fd30f27ffeb3c3780d0b84f496518868445c7a8db003639a649cc98453"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:7a09237a795d11cd11f9dae505d170a29b5616151db1e10c14f892b11caadc7d"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:de01c9a3a3b7b69627d624ff69d9f11d28ce9908eea2fb6245adafa4b1d43df6"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8c1f25c5069ee42a5bcffdc66c300a407941edd73f3239e9fdefd26216407391"}, + {file = "safetensors-0.4.2-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:7a73b3649456d09ca8506140d44484b63154a7378434cc1e8719f8056550b224"}, + {file = "safetensors-0.4.2-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:e1625a8d07d046e968bd5c4961810aba1225984e4fb9243626f9d04a06ed3fee"}, + {file = "safetensors-0.4.2-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:8f74c86b25615cb24ad4cff765a2eefc09d71bf0fed97588cf585aad9c38fbb4"}, + {file = "safetensors-0.4.2-cp312-none-win32.whl", hash = "sha256:8523b9c5777d771bcde5c2389c03f1cdf7ebe8797432a1bd5e345efe25c55987"}, + {file = "safetensors-0.4.2-cp312-none-win_amd64.whl", hash = "sha256:dcff0243e1737a21f83d664c63fed89d1f532c23fc6830d0427279fabd789ccb"}, + {file = "safetensors-0.4.2-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:96ad3d7d472612e26cbe413922b4fb13933310f0511d346ea5cc9a1e856e52eb"}, + {file = "safetensors-0.4.2-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:88250922401b5ae4e37de929178caf46be47ed16c817b2237b81679bec07c120"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d40443554142fc0ab30652d5cc8554c4b7a613513bde00373e18afd5de8cbe4b"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:27f53f70106224d32d874aacecbeb4a6e4c5b16a1d2006d0e876d97229086d71"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:cc068afe23734dfb26ce19db0a7877499ddf73b1d55ceb762417e8da4a1b05fb"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:9be1918eb8d43a11a6f8806759fccfa0eeb0542b12924caba66af8a7800ad01a"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:41911087d20a7bbd78cb4ad4f98aab0c431533107584df6635d8b54b99945573"}, + {file = "safetensors-0.4.2-cp313-cp313-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:50771c662aab909f31e94d048e76861fd027d66076ea773eef2e66c717766e24"}, + {file = "safetensors-0.4.2-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:13f2e57be007b7ea9329133d2399e6bdfcf1910f655440a4da17df3a45afcd30"}, + {file = "safetensors-0.4.2-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:c772147e6395bc829842e0a98e1b30c67fe25d816299c28196488511d5a5e951"}, + {file = "safetensors-0.4.2-cp37-cp37m-macosx_10_12_x86_64.whl", hash = "sha256:36239a0060b537a3e8c473df78cffee14c3ec4f51d5f1a853af99371a2fb2a35"}, + {file = "safetensors-0.4.2-cp37-cp37m-macosx_11_0_arm64.whl", hash = "sha256:d0cbb7664fad2c307f95195f951b7059e95dc23e0e1822e5978c8b500098543c"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:2b3e55adb6bd9dc1c2a341e72f48f075953fa35d173dd8e29a95b3b02d0d1462"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:42f743b3cca863fba53ca57a193f510e5ec359b97f38c282437716b6768e4a25"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:04e6af4a6dbeb06c4e6e7d46cf9c716cbc4cc5ef62584fd8a7c0fe558562df45"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a492ba21b5c8f14ee5ec9b20f42ba969e53ca1f909a4d04aad736b66a341dcc2"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b25b8233a1a85dc67e39838951cfb01595d792f3b7b644add63edb652992e030"}, + {file = "safetensors-0.4.2-cp37-cp37m-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:fd27e063fbdafe776f7b1714da59110e88f270e86db00788a8fd65f4eacfeba7"}, + {file = "safetensors-0.4.2-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:1b6fa399f251bbeb52029bf5a0ac2878d7705dd3612a2f8895b48e9c11f0367d"}, + {file = "safetensors-0.4.2-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:de642d46b459e4afd5c2020b26c0d6d869a171ea00411897d5776c127cac74f0"}, + {file = "safetensors-0.4.2-cp37-none-win32.whl", hash = "sha256:77b72d17754c93bb68f3598182f14d78776e0b9b31682ca5bb2c7c5bd9a75267"}, + {file = "safetensors-0.4.2-cp37-none-win_amd64.whl", hash = "sha256:d36ee3244d461cd655aeef493792c3bccf4875282f8407fd9af99e9a41cf2530"}, + {file = "safetensors-0.4.2-cp38-cp38-macosx_10_12_x86_64.whl", hash = "sha256:16b6b3884f7876c6b3b23a742428223a7170a5a9dac819d8c12a1569422c4b5a"}, + {file = "safetensors-0.4.2-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:ee25d311493fbbe0be9d395faee46e9d79e8948f461e388ff39e59875ed9a350"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:eed8097968585cd752a1171f86fce9aa1d89a29033e5cd8bec5a502e29f6b7af"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:880e6865cf72cb67f9ab8d04a3c4b49dd95ae92fb1583929ce65aed94e1f685f"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:91290f83daf80ce6d1a7f629b244443c200060a80f908b29d879021409e5ea94"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3517d568486ab3508a7acc360b82d7a4a3e26b86efdf210a9ecd9d233c40708a"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e1f43a77eb38540f782999e5dc5645164fe9027d3f0194f6c9a5126168017efa"}, + {file = "safetensors-0.4.2-cp38-cp38-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:b684d9818aa5d63fddc65f7d0151968037d255d91adf74eba82125b41c680aaa"}, + {file = "safetensors-0.4.2-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:ab1f5d84185f9fefaf21413efb764e4908057b8a9a0b987ede890c353490fd70"}, + {file = "safetensors-0.4.2-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:2bd979642e6c3a517ef4b84ff36c2fee4015664fea05a61154fc565978347553"}, + {file = "safetensors-0.4.2-cp38-none-win32.whl", hash = "sha256:11be6e7afed29e5a5628f0aa6214e34bc194da73f558dc69fc7d56e07037422a"}, + {file = "safetensors-0.4.2-cp38-none-win_amd64.whl", hash = "sha256:2f7a6e5d29bd2cc340cffaa391fa437b1be9d21a2bd8b8724d2875d13a6ef2a9"}, + {file = "safetensors-0.4.2-cp39-cp39-macosx_10_12_x86_64.whl", hash = "sha256:a5a921b4fe6925f9942adff3ebae8c16e0487908c54586a5a42f35b59fd69794"}, + {file = "safetensors-0.4.2-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:b691727228c28f2d82d8a92b2bc26e7a1f129ee40b2f2a3185b5974e038ed47c"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:91ca1056decc4e981248786e87b2a202d4841ee5f99d433f1adf3d44d4bcfa0e"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:55969fd2e6fdb38dc221b0ab380668c21b0efa12a7562db9924759faa3c51757"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:6ae429bfaecc10ab5fe78c93009b3d1656c1581da560041e700eadb497dbe7a4"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4ff88f194fe4ac50b463a4a6f0c03af9ad72eb5d24ec6d6730af59522e37fedb"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a80cb48d0a447f8dd18e61813efa7d3f8f8d52edf0f05806abc0c59b83431f57"}, + {file = "safetensors-0.4.2-cp39-cp39-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:b286fb7adfee70a4189898ac2342b8a67d5f493e6b21b0af89ca8eac1b967cbf"}, + {file = "safetensors-0.4.2-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:0ceeff9ddbab4f78738489eb6682867ae946178776f33699737b2129b5394dc1"}, + {file = "safetensors-0.4.2-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:a26fae748a7488cb3aac381eddfa818c42052c87b5e689fb4c6e82ed58cec209"}, + {file = "safetensors-0.4.2-cp39-none-win32.whl", hash = "sha256:039a42ab33c9d68b39706fd38f1922ace26866eff246bf20271edb619f5f848b"}, + {file = "safetensors-0.4.2-cp39-none-win_amd64.whl", hash = "sha256:b3a3e1f5b85859e398773f064943b62a4059f225008a2a8ee6add1edcf77cacf"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-macosx_10_12_x86_64.whl", hash = "sha256:4e70d442ad17e8b153ef9095bf48ea64f15a66bf26dc2b6ca94660c154edbc24"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-macosx_11_0_arm64.whl", hash = "sha256:b90f1d9809caf4ff395951b4703295a68d12907f6945bbc3129e934ff8ae46f6"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8c7ac9ad3728838006598e296b3ae9f27d80b489effd4685b92d97b3fc4c98f6"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:de5730d77e6ff7f4c7039e20913661ad0ea2f86c09e71c039e73dfdd1f394f08"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:44feb8cb156d6803dcd19fc6b81b27235f29b877660605a6ac35e1da7d64f0e4"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:523a241c33e7c827ab9a3a23760d75c7d062f43dfe55b6b019409f89b0fb52d1"}, + {file = "safetensors-0.4.2-pp310-pypy310_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:fb18300e8eb74291225214f26c9a8ae2110fd61a6c9b5a2ff4c4e0eb1bb9a998"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-macosx_10_12_x86_64.whl", hash = "sha256:fe5437ff9fb116e44f2ab558981249ae63f978392b4576e62fcfe167d353edbc"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d9304a0934ced5a5d272f39de36291dc141dfc152d277f03fb4d65f2fb2ffa7c"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:160ba1b1e11cf874602c233ab80a14f588571d09556cbc3586900121d622b5ed"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:04fcd6fcf7d9c13c7e5dc7e08de5e492ee4daa8f4ad74b4d8299d3eb0224292f"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:906d14c4a677d35834fb0f3a5455ef8305e1bba10a5e0f2e0f357b3d1ad989f2"}, + {file = "safetensors-0.4.2-pp37-pypy37_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:df3fcdec0cd543084610d1f09c65cdb10fb3079f79bceddc092b0d187c6a265b"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-macosx_10_12_x86_64.whl", hash = "sha256:5ca76f13fb1cef242ea3ad2cb37388e7d005994f42af8b44bee56ba48b2d45ce"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-macosx_11_0_arm64.whl", hash = "sha256:278a1a3414c020785decdcd741c578725721274d2f9f787fcc930882e83b89cc"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:05b5a461cc68ecd42d9d546e5e1268a39d8ede7934a68d1ce17c3c659cb829d6"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c2341411412a41671d25e26bed59ec121e46bf4fadb8132895e610411c4b9681"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:3497ac3895acf17c5f98197f1fa4769f09c5e7ede07fcb102f1c201e663e052c"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:01b5e71d3754d2201294f1eb7a6d59cce3a5702ff96d83d226571b2ca2183837"}, + {file = "safetensors-0.4.2-pp38-pypy38_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:3627dbd1ea488dd8046a0491de5087f3c0d641e7acc80c0189a33c69398f1cd1"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-macosx_10_12_x86_64.whl", hash = "sha256:9d56f0ef53afad26ec54ceede78a43e9a23a076dadbbda7b44d304c591abf4c1"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-macosx_11_0_arm64.whl", hash = "sha256:b259ca73d42daf658a1bda463f1f83885ae4d93a60869be80d7f7dfcc9d8bbb5"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:1ebc3cd401e4eb54e7c0a70346be565e81942d9a41fafd5f4bf7ab3a55d10378"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:5bc384a0309b706aa0425c93abb0390508a61bf029ce99c7d9df4220f25871a5"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:af2d8f7235d8a08fbccfb8394387890e7fa38942b349a94e6eff13c52ac98087"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:0911315bbcc5289087d063c2c2c7ccd711ea97a7e557a7bce005ac2cf80146aa"}, + {file = "safetensors-0.4.2-pp39-pypy39_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:1efe31673be91832d73439a2af426743e1395fc9ef7b081914e9e1d567bd7b5f"}, + {file = "safetensors-0.4.2.tar.gz", hash = "sha256:acc85dcb09ec5e8aa787f588d7ad4d55c103f31e4ff060e17d92cc0e8b8cac73"}, +] + +[package.extras] +all = ["safetensors[jax]", "safetensors[numpy]", "safetensors[paddlepaddle]", "safetensors[pinned-tf]", "safetensors[quality]", "safetensors[testing]", "safetensors[torch]"] +dev = ["safetensors[all]"] +jax = ["flax (>=0.6.3)", "jax (>=0.3.25)", "jaxlib (>=0.3.25)", "safetensors[numpy]"] +mlx = ["mlx (>=0.0.9)"] +numpy = ["numpy (>=1.21.6)"] +paddlepaddle = ["paddlepaddle (>=2.4.1)", "safetensors[numpy]"] +pinned-tf = ["safetensors[numpy]", "tensorflow (==2.11.0)"] +quality = ["black (==22.3)", "click (==8.0.4)", "flake8 (>=3.8.3)", "isort (>=5.5.4)"] +tensorflow = ["safetensors[numpy]", "tensorflow (>=2.11.0)"] +testing = ["h5py (>=3.7.0)", "huggingface_hub (>=0.12.1)", "hypothesis (>=6.70.2)", "pytest (>=7.2.0)", "pytest-benchmark (>=4.0.0)", "safetensors[numpy]", "setuptools_rust (>=1.5.2)"] +torch = ["safetensors[numpy]", "torch (>=1.10)"] + +[[package]] +name = "sentencepiece" +version = "0.1.99" +description = "SentencePiece python wrapper" +optional = false +python-versions = "*" +files = [ + {file = "sentencepiece-0.1.99-cp310-cp310-macosx_10_9_universal2.whl", hash = "sha256:0eb528e70571b7c02723e5804322469b82fe7ea418c96051d0286c0fa028db73"}, + {file = "sentencepiece-0.1.99-cp310-cp310-macosx_10_9_x86_64.whl", hash = "sha256:77d7fafb2c4e4659cbdf303929503f37a26eabc4ff31d3a79bf1c5a1b338caa7"}, + {file = "sentencepiece-0.1.99-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:be9cf5b9e404c245aeb3d3723c737ba7a8f5d4ba262ef233a431fa6c45f732a0"}, + {file = "sentencepiece-0.1.99-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:baed1a26464998f9710d20e52607c29ffd4293e7c71c6a1f83f51ad0911ec12c"}, + {file = "sentencepiece-0.1.99-cp310-cp310-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:9832f08bb372d4c8b567612f8eab9e36e268dff645f1c28f9f8e851be705f6d1"}, + {file = "sentencepiece-0.1.99-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:019e7535108e309dae2b253a75834fc3128240aa87c00eb80732078cdc182588"}, + {file = "sentencepiece-0.1.99-cp310-cp310-win32.whl", hash = "sha256:fa16a830416bb823fa2a52cbdd474d1f7f3bba527fd2304fb4b140dad31bb9bc"}, + {file = "sentencepiece-0.1.99-cp310-cp310-win_amd64.whl", hash = "sha256:14b0eccb7b641d4591c3e12ae44cab537d68352e4d3b6424944f0c447d2348d5"}, + {file = "sentencepiece-0.1.99-cp311-cp311-macosx_10_9_universal2.whl", hash = "sha256:6d3c56f24183a1e8bd61043ff2c58dfecdc68a5dd8955dc13bab83afd5f76b81"}, + {file = "sentencepiece-0.1.99-cp311-cp311-macosx_10_9_x86_64.whl", hash = "sha256:ed6ea1819fd612c989999e44a51bf556d0ef6abfb553080b9be3d347e18bcfb7"}, + {file = "sentencepiece-0.1.99-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:a2a0260cd1fb7bd8b4d4f39dc2444a8d5fd4e0a0c4d5c899810ef1abf99b2d45"}, + {file = "sentencepiece-0.1.99-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:8a1abff4d1ff81c77cac3cc6fefa34fa4b8b371e5ee51cb7e8d1ebc996d05983"}, + {file = "sentencepiece-0.1.99-cp311-cp311-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:004e6a621d4bc88978eecb6ea7959264239a17b70f2cbc348033d8195c9808ec"}, + {file = "sentencepiece-0.1.99-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:db361e03342c41680afae5807590bc88aa0e17cfd1a42696a160e4005fcda03b"}, + {file = "sentencepiece-0.1.99-cp311-cp311-win32.whl", hash = "sha256:2d95e19168875b70df62916eb55428a0cbcb834ac51d5a7e664eda74def9e1e0"}, + {file = "sentencepiece-0.1.99-cp311-cp311-win_amd64.whl", hash = "sha256:f90d73a6f81248a909f55d8e6ef56fec32d559e1e9af045f0b0322637cb8e5c7"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-macosx_10_9_x86_64.whl", hash = "sha256:62e24c81e74bd87a6e0d63c51beb6527e4c0add67e1a17bac18bcd2076afcfeb"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:57efcc2d51caff20d9573567d9fd3f854d9efe613ed58a439c78c9f93101384a"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:6a904c46197993bd1e95b93a6e373dca2f170379d64441041e2e628ad4afb16f"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:d89adf59854741c0d465f0e1525b388c0d174f611cc04af54153c5c4f36088c4"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-win32.whl", hash = "sha256:47c378146928690d1bc106fdf0da768cebd03b65dd8405aa3dd88f9c81e35dba"}, + {file = "sentencepiece-0.1.99-cp36-cp36m-win_amd64.whl", hash = "sha256:9ba142e7a90dd6d823c44f9870abdad45e6c63958eb60fe44cca6828d3b69da2"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-macosx_10_9_x86_64.whl", hash = "sha256:b7b1a9ae4d7c6f1f867e63370cca25cc17b6f4886729595b885ee07a58d3cec3"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d0f644c9d4d35c096a538507b2163e6191512460035bf51358794a78515b74f7"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:c8843d23a0f686d85e569bd6dcd0dd0e0cbc03731e63497ca6d5bacd18df8b85"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:33e6f690a1caebb4867a2e367afa1918ad35be257ecdb3455d2bbd787936f155"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-win32.whl", hash = "sha256:8a321866c2f85da7beac74a824b4ad6ddc2a4c9bccd9382529506d48f744a12c"}, + {file = "sentencepiece-0.1.99-cp37-cp37m-win_amd64.whl", hash = "sha256:c42f753bcfb7661c122a15b20be7f684b61fc8592c89c870adf52382ea72262d"}, + {file = "sentencepiece-0.1.99-cp38-cp38-macosx_10_9_universal2.whl", hash = "sha256:85b476406da69c70586f0bb682fcca4c9b40e5059814f2db92303ea4585c650c"}, + {file = "sentencepiece-0.1.99-cp38-cp38-macosx_10_9_x86_64.whl", hash = "sha256:cfbcfe13c69d3f87b7fcd5da168df7290a6d006329be71f90ba4f56bc77f8561"}, + {file = "sentencepiece-0.1.99-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:445b0ec381af1cd4eef95243e7180c63d9c384443c16c4c47a28196bd1cda937"}, + {file = "sentencepiece-0.1.99-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c6890ea0f2b4703f62d0bf27932e35808b1f679bdb05c7eeb3812b935ba02001"}, + {file = "sentencepiece-0.1.99-cp38-cp38-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:fb71af492b0eefbf9f2501bec97bcd043b6812ab000d119eaf4bd33f9e283d03"}, + {file = "sentencepiece-0.1.99-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:27b866b5bd3ddd54166bbcbf5c8d7dd2e0b397fac8537991c7f544220b1f67bc"}, + {file = "sentencepiece-0.1.99-cp38-cp38-win32.whl", hash = "sha256:b133e8a499eac49c581c3c76e9bdd08c338cc1939e441fee6f92c0ccb5f1f8be"}, + {file = "sentencepiece-0.1.99-cp38-cp38-win_amd64.whl", hash = "sha256:0eaf3591dd0690a87f44f4df129cf8d05d8a4029b5b6709b489b8e27f9a9bcff"}, + {file = "sentencepiece-0.1.99-cp39-cp39-macosx_10_9_universal2.whl", hash = "sha256:38efeda9bbfb55052d482a009c6a37e52f42ebffcea9d3a98a61de7aee356a28"}, + {file = "sentencepiece-0.1.99-cp39-cp39-macosx_10_9_x86_64.whl", hash = "sha256:6c030b081dc1e1bcc9fadc314b19b740715d3d566ad73a482da20d7d46fd444c"}, + {file = "sentencepiece-0.1.99-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:84dbe53e02e4f8a2e45d2ac3e430d5c83182142658e25edd76539b7648928727"}, + {file = "sentencepiece-0.1.99-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:0b0f55d0a0ee1719b4b04221fe0c9f0c3461dc3dabd77a035fa2f4788eb3ef9a"}, + {file = "sentencepiece-0.1.99-cp39-cp39-manylinux_2_17_i686.manylinux2014_i686.whl", hash = "sha256:18e800f206cd235dc27dc749299e05853a4e4332e8d3dfd81bf13d0e5b9007d9"}, + {file = "sentencepiece-0.1.99-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:2ae1c40cda8f9d5b0423cfa98542735c0235e7597d79caf318855cdf971b2280"}, + {file = "sentencepiece-0.1.99-cp39-cp39-win32.whl", hash = "sha256:c84ce33af12ca222d14a1cdd37bd76a69401e32bc68fe61c67ef6b59402f4ab8"}, + {file = "sentencepiece-0.1.99-cp39-cp39-win_amd64.whl", hash = "sha256:350e5c74d739973f1c9643edb80f7cc904dc948578bcb1d43c6f2b173e5d18dd"}, + {file = "sentencepiece-0.1.99.tar.gz", hash = "sha256:189c48f5cb2949288f97ccdb97f0473098d9c3dcf5a3d99d4eabe719ec27297f"}, +] + +[[package]] +name = "sympy" +version = "1.12" +description = "Computer algebra system (CAS) in Python" +optional = false +python-versions = ">=3.8" +files = [ + {file = "sympy-1.12-py3-none-any.whl", hash = "sha256:c3588cd4295d0c0f603d0f2ae780587e64e2efeedb3521e46b9bb1d08d184fa5"}, + {file = "sympy-1.12.tar.gz", hash = "sha256:ebf595c8dac3e0fdc4152c51878b498396ec7f30e7a914d6071e674d49420fb8"}, +] + +[package.dependencies] +mpmath = ">=0.19" + +[[package]] +name = "tokenizers" +version = "0.15.2" +description = "" +optional = false +python-versions = ">=3.7" +files = [ + {file = "tokenizers-0.15.2-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:52f6130c9cbf70544287575a985bf44ae1bda2da7e8c24e97716080593638012"}, + {file = "tokenizers-0.15.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:054c1cc9c6d68f7ffa4e810b3d5131e0ba511b6e4be34157aa08ee54c2f8d9ee"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:a9b9b070fdad06e347563b88c278995735292ded1132f8657084989a4c84a6d5"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:ea621a7eef4b70e1f7a4e84dd989ae3f0eeb50fc8690254eacc08acb623e82f1"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:cf7fd9a5141634fa3aa8d6b7be362e6ae1b4cda60da81388fa533e0b552c98fd"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:44f2a832cd0825295f7179eaf173381dc45230f9227ec4b44378322d900447c9"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:8b9ec69247a23747669ec4b0ca10f8e3dfb3545d550258129bd62291aabe8605"}, + {file = "tokenizers-0.15.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:40b6a4c78da863ff26dbd5ad9a8ecc33d8a8d97b535172601cf00aee9d7ce9ce"}, + {file = "tokenizers-0.15.2-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:5ab2a4d21dcf76af60e05af8063138849eb1d6553a0d059f6534357bce8ba364"}, + {file = "tokenizers-0.15.2-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:a47acfac7e511f6bbfcf2d3fb8c26979c780a91e06fb5b9a43831b2c0153d024"}, + {file = "tokenizers-0.15.2-cp310-none-win32.whl", hash = "sha256:064ff87bb6acdbd693666de9a4b692add41308a2c0ec0770d6385737117215f2"}, + {file = "tokenizers-0.15.2-cp310-none-win_amd64.whl", hash = "sha256:3b919afe4df7eb6ac7cafd2bd14fb507d3f408db7a68c43117f579c984a73843"}, + {file = "tokenizers-0.15.2-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:89cd1cb93e4b12ff39bb2d626ad77e35209de9309a71e4d3d4672667b4b256e7"}, + {file = "tokenizers-0.15.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:cfed5c64e5be23d7ee0f0e98081a25c2a46b0b77ce99a4f0605b1ec43dd481fa"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:a907d76dcfda37023ba203ab4ceeb21bc5683436ebefbd895a0841fd52f6f6f2"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:20ea60479de6fc7b8ae756b4b097572372d7e4032e2521c1bbf3d90c90a99ff0"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:48e2b9335be2bc0171df9281385c2ed06a15f5cf121c44094338306ab7b33f2c"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:112a1dd436d2cc06e6ffdc0b06d55ac019a35a63afd26475205cb4b1bf0bfbff"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4620cca5c2817177ee8706f860364cc3a8845bc1e291aaf661fb899e5d1c45b0"}, + {file = "tokenizers-0.15.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ccd73a82751c523b3fc31ff8194702e4af4db21dc20e55b30ecc2079c5d43cb7"}, + {file = "tokenizers-0.15.2-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:107089f135b4ae7817affe6264f8c7a5c5b4fd9a90f9439ed495f54fcea56fb4"}, + {file = "tokenizers-0.15.2-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:0ff110ecc57b7aa4a594396525a3451ad70988e517237fe91c540997c4e50e29"}, + {file = "tokenizers-0.15.2-cp311-none-win32.whl", hash = "sha256:6d76f00f5c32da36c61f41c58346a4fa7f0a61be02f4301fd30ad59834977cc3"}, + {file = "tokenizers-0.15.2-cp311-none-win_amd64.whl", hash = "sha256:cc90102ed17271cf0a1262babe5939e0134b3890345d11a19c3145184b706055"}, + {file = "tokenizers-0.15.2-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:f86593c18d2e6248e72fb91c77d413a815153b8ea4e31f7cd443bdf28e467670"}, + {file = "tokenizers-0.15.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:0774bccc6608eca23eb9d620196687c8b2360624619623cf4ba9dc9bd53e8b51"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:d0222c5b7c9b26c0b4822a82f6a7011de0a9d3060e1da176f66274b70f846b98"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:3835738be1de66624fff2f4f6f6684775da4e9c00bde053be7564cbf3545cc66"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:0143e7d9dcd811855c1ce1ab9bf5d96d29bf5e528fd6c7824d0465741e8c10fd"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:db35825f6d54215f6b6009a7ff3eedee0848c99a6271c870d2826fbbedf31a38"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:3f5e64b0389a2be47091d8cc53c87859783b837ea1a06edd9d8e04004df55a5c"}, + {file = "tokenizers-0.15.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:9e0480c452217edd35eca56fafe2029fb4d368b7c0475f8dfa3c5c9c400a7456"}, + {file = "tokenizers-0.15.2-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:a33ab881c8fe70474980577e033d0bc9a27b7ab8272896e500708b212995d834"}, + {file = "tokenizers-0.15.2-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:a308a607ca9de2c64c1b9ba79ec9a403969715a1b8ba5f998a676826f1a7039d"}, + {file = "tokenizers-0.15.2-cp312-none-win32.whl", hash = "sha256:b8fcfa81bcb9447df582c5bc96a031e6df4da2a774b8080d4f02c0c16b42be0b"}, + {file = "tokenizers-0.15.2-cp312-none-win_amd64.whl", hash = "sha256:38d7ab43c6825abfc0b661d95f39c7f8af2449364f01d331f3b51c94dcff7221"}, + {file = "tokenizers-0.15.2-cp313-cp313-macosx_10_12_x86_64.whl", hash = "sha256:38bfb0204ff3246ca4d5e726e8cc8403bfc931090151e6eede54d0e0cf162ef0"}, + {file = "tokenizers-0.15.2-cp313-cp313-macosx_11_0_arm64.whl", hash = "sha256:9c861d35e8286a53e06e9e28d030b5a05bcbf5ac9d7229e561e53c352a85b1fc"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:936bf3842db5b2048eaa53dade907b1160f318e7c90c74bfab86f1e47720bdd6"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:620beacc3373277700d0e27718aa8b25f7b383eb8001fba94ee00aeea1459d89"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:2735ecbbf37e52db4ea970e539fd2d450d213517b77745114f92867f3fc246eb"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:473c83c5e2359bb81b0b6fde870b41b2764fcdd36d997485e07e72cc3a62264a"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:968fa1fb3c27398b28a4eca1cbd1e19355c4d3a6007f7398d48826bbe3a0f728"}, + {file = "tokenizers-0.15.2-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:865c60ae6eaebdde7da66191ee9b7db52e542ed8ee9d2c653b6d190a9351b980"}, + {file = "tokenizers-0.15.2-cp313-cp313-musllinux_1_1_aarch64.whl", hash = "sha256:7c0d8b52664ab2d4a8d6686eb5effc68b78608a9008f086a122a7b2996befbab"}, + {file = "tokenizers-0.15.2-cp313-cp313-musllinux_1_1_x86_64.whl", hash = "sha256:f33dfbdec3784093a9aebb3680d1f91336c56d86cc70ddf88708251da1fe9064"}, + {file = "tokenizers-0.15.2-cp37-cp37m-macosx_10_12_x86_64.whl", hash = "sha256:d44ba80988ff9424e33e0a49445072ac7029d8c0e1601ad25a0ca5f41ed0c1d6"}, + {file = "tokenizers-0.15.2-cp37-cp37m-macosx_11_0_arm64.whl", hash = "sha256:dce74266919b892f82b1b86025a613956ea0ea62a4843d4c4237be2c5498ed3a"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:0ef06b9707baeb98b316577acb04f4852239d856b93e9ec3a299622f6084e4be"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c73e2e74bbb07910da0d37c326869f34113137b23eadad3fc00856e6b3d9930c"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:4eeb12daf02a59e29f578a865f55d87cd103ce62bd8a3a5874f8fdeaa82e336b"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:9ba9f6895af58487ca4f54e8a664a322f16c26bbb442effd01087eba391a719e"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ccec77aa7150e38eec6878a493bf8c263ff1fa8a62404e16c6203c64c1f16a26"}, + {file = "tokenizers-0.15.2-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:f3f40604f5042ff210ba82743dda2b6aa3e55aa12df4e9f2378ee01a17e2855e"}, + {file = "tokenizers-0.15.2-cp37-cp37m-musllinux_1_1_aarch64.whl", hash = "sha256:5645938a42d78c4885086767c70923abad047163d809c16da75d6b290cb30bbe"}, + {file = "tokenizers-0.15.2-cp37-cp37m-musllinux_1_1_x86_64.whl", hash = "sha256:05a77cbfebe28a61ab5c3891f9939cc24798b63fa236d84e5f29f3a85a200c00"}, + {file = "tokenizers-0.15.2-cp37-none-win32.whl", hash = "sha256:361abdc068e8afe9c5b818769a48624687fb6aaed49636ee39bec4e95e1a215b"}, + {file = "tokenizers-0.15.2-cp37-none-win_amd64.whl", hash = "sha256:7ef789f83eb0f9baeb4d09a86cd639c0a5518528f9992f38b28e819df397eb06"}, + {file = "tokenizers-0.15.2-cp38-cp38-macosx_10_12_x86_64.whl", hash = "sha256:4fe1f74a902bee74a3b25aff180fbfbf4f8b444ab37c4d496af7afd13a784ed2"}, + {file = "tokenizers-0.15.2-cp38-cp38-macosx_11_0_arm64.whl", hash = "sha256:4c4b89038a684f40a6b15d6b09f49650ac64d951ad0f2a3ea9169687bbf2a8ba"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:d05a1b06f986d41aed5f2de464c003004b2df8aaf66f2b7628254bcbfb72a438"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:508711a108684111ec8af89d3a9e9e08755247eda27d0ba5e3c50e9da1600f6d"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:daa348f02d15160cb35439098ac96e3a53bacf35885072611cd9e5be7d333daa"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:494fdbe5932d3416de2a85fc2470b797e6f3226c12845cadf054dd906afd0442"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:c2d60f5246f4da9373f75ff18d64c69cbf60c3bca597290cea01059c336d2470"}, + {file = "tokenizers-0.15.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:93268e788825f52de4c7bdcb6ebc1fcd4a5442c02e730faa9b6b08f23ead0e24"}, + {file = "tokenizers-0.15.2-cp38-cp38-musllinux_1_1_aarch64.whl", hash = "sha256:6fc7083ab404019fc9acafe78662c192673c1e696bd598d16dc005bd663a5cf9"}, + {file = "tokenizers-0.15.2-cp38-cp38-musllinux_1_1_x86_64.whl", hash = "sha256:41e39b41e5531d6b2122a77532dbea60e171ef87a3820b5a3888daa847df4153"}, + {file = "tokenizers-0.15.2-cp38-none-win32.whl", hash = "sha256:06cd0487b1cbfabefb2cc52fbd6b1f8d4c37799bd6c6e1641281adaa6b2504a7"}, + {file = "tokenizers-0.15.2-cp38-none-win_amd64.whl", hash = "sha256:5179c271aa5de9c71712e31cb5a79e436ecd0d7532a408fa42a8dbfa4bc23fd9"}, + {file = "tokenizers-0.15.2-cp39-cp39-macosx_10_12_x86_64.whl", hash = "sha256:82f8652a74cc107052328b87ea8b34291c0f55b96d8fb261b3880216a9f9e48e"}, + {file = "tokenizers-0.15.2-cp39-cp39-macosx_11_0_arm64.whl", hash = "sha256:02458bee6f5f3139f1ebbb6d042b283af712c0981f5bc50edf771d6b762d5e4f"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:c9a09cd26cca2e1c349f91aa665309ddb48d71636370749414fbf67bc83c5343"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:158be8ea8554e5ed69acc1ce3fbb23a06060bd4bbb09029431ad6b9a466a7121"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:1ddba9a2b0c8c81633eca0bb2e1aa5b3a15362b1277f1ae64176d0f6eba78ab1"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:3ef5dd1d39797044642dbe53eb2bc56435308432e9c7907728da74c69ee2adca"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:454c203164e07a860dbeb3b1f4a733be52b0edbb4dd2e5bd75023ffa8b49403a"}, + {file = "tokenizers-0.15.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0cf6b7f1d4dc59af960e6ffdc4faffe6460bbfa8dce27a58bf75755ffdb2526d"}, + {file = "tokenizers-0.15.2-cp39-cp39-musllinux_1_1_aarch64.whl", hash = "sha256:2ef09bbc16519f6c25d0c7fc0c6a33a6f62923e263c9d7cca4e58b8c61572afb"}, + {file = "tokenizers-0.15.2-cp39-cp39-musllinux_1_1_x86_64.whl", hash = "sha256:c9a2ebdd2ad4ec7a68e7615086e633857c85e2f18025bd05d2a4399e6c5f7169"}, + {file = "tokenizers-0.15.2-cp39-none-win32.whl", hash = "sha256:918fbb0eab96fe08e72a8c2b5461e9cce95585d82a58688e7f01c2bd546c79d0"}, + {file = "tokenizers-0.15.2-cp39-none-win_amd64.whl", hash = "sha256:524e60da0135e106b254bd71f0659be9f89d83f006ea9093ce4d1fab498c6d0d"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-macosx_10_12_x86_64.whl", hash = "sha256:6a9b648a58281c4672212fab04e60648fde574877d0139cd4b4f93fe28ca8944"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-macosx_11_0_arm64.whl", hash = "sha256:7c7d18b733be6bbca8a55084027f7be428c947ddf871c500ee603e375013ffba"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:13ca3611de8d9ddfbc4dc39ef54ab1d2d4aaa114ac8727dfdc6a6ec4be017378"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:237d1bf3361cf2e6463e6c140628e6406766e8b27274f5fcc62c747ae3c6f094"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:67a0fe1e49e60c664915e9fb6b0cb19bac082ab1f309188230e4b2920230edb3"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:4e022fe65e99230b8fd89ebdfea138c24421f91c1a4f4781a8f5016fd5cdfb4d"}, + {file = "tokenizers-0.15.2-pp310-pypy310_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:d857be2df69763362ac699f8b251a8cd3fac9d21893de129bc788f8baaef2693"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-macosx_10_12_x86_64.whl", hash = "sha256:708bb3e4283177236309e698da5fcd0879ce8fd37457d7c266d16b550bcbbd18"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:64c35e09e9899b72a76e762f9854e8750213f67567787d45f37ce06daf57ca78"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:c1257f4394be0d3b00de8c9e840ca5601d0a4a8438361ce9c2b05c7d25f6057b"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:02272fe48280e0293a04245ca5d919b2c94a48b408b55e858feae9618138aeda"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:dc3ad9ebc76eabe8b1d7c04d38be884b8f9d60c0cdc09b0aa4e3bcf746de0388"}, + {file = "tokenizers-0.15.2-pp37-pypy37_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:32e16bdeffa7c4f46bf2152172ca511808b952701d13e7c18833c0b73cb5c23f"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-macosx_10_12_x86_64.whl", hash = "sha256:fb16ba563d59003028b678d2361a27f7e4ae0ab29c7a80690efa20d829c81fdb"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-macosx_11_0_arm64.whl", hash = "sha256:2277c36d2d6cdb7876c274547921a42425b6810d38354327dd65a8009acf870c"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:1cf75d32e8d250781940d07f7eece253f2fe9ecdb1dc7ba6e3833fa17b82fcbc"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:f1b3b31884dc8e9b21508bb76da80ebf7308fdb947a17affce815665d5c4d028"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b10122d8d8e30afb43bb1fe21a3619f62c3e2574bff2699cf8af8b0b6c5dc4a3"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:d88b96ff0fe8e91f6ef01ba50b0d71db5017fa4e3b1d99681cec89a85faf7bf7"}, + {file = "tokenizers-0.15.2-pp38-pypy38_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:37aaec5a52e959892870a7c47cef80c53797c0db9149d458460f4f31e2fb250e"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-macosx_10_12_x86_64.whl", hash = "sha256:e2ea752f2b0fe96eb6e2f3adbbf4d72aaa1272079b0dfa1145507bd6a5d537e6"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-macosx_11_0_arm64.whl", hash = "sha256:4b19a808d8799fda23504a5cd31d2f58e6f52f140380082b352f877017d6342b"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-manylinux_2_12_i686.manylinux2010_i686.whl", hash = "sha256:64c86e5e068ac8b19204419ed8ca90f9d25db20578f5881e337d203b314f4104"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:de19c4dc503c612847edf833c82e9f73cd79926a384af9d801dcf93f110cea4e"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:ea09acd2fe3324174063d61ad620dec3bcf042b495515f27f638270a7d466e8b"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-musllinux_1_1_aarch64.whl", hash = "sha256:cf27fd43472e07b57cf420eee1e814549203d56de00b5af8659cb99885472f1f"}, + {file = "tokenizers-0.15.2-pp39-pypy39_pp73-musllinux_1_1_x86_64.whl", hash = "sha256:7ca22bd897537a0080521445d91a58886c8c04084a6a19e6c78c586e0cfa92a5"}, + {file = "tokenizers-0.15.2.tar.gz", hash = "sha256:e6e9c6e019dd5484be5beafc775ae6c925f4c69a3487040ed09b45e13df2cb91"}, +] + +[package.dependencies] +huggingface_hub = ">=0.16.4,<1.0" + +[package.extras] +dev = ["tokenizers[testing]"] +docs = ["setuptools_rust", "sphinx", "sphinx_rtd_theme"] +testing = ["black (==22.3)", "datasets", "numpy", "pytest", "requests"] + +[[package]] +name = "torch" +version = "2.2.1+cpu" +description = "Tensors and Dynamic neural networks in Python with strong GPU acceleration" +optional = false +python-versions = ">=3.8.0" +files = [ + {file = "torch-2.2.1+cpu-cp310-cp310-linux_x86_64.whl", hash = "sha256:5d82422cf04797f1b2a8574b64a916070ec83eef58ad4900615ee0218d7b8b8e"}, + {file = "torch-2.2.1+cpu-cp310-cp310-win_amd64.whl", hash = "sha256:f8914dd0f5f0e5c66fdecd9559403eea9feac82d1ea639b672fde0073c6addbd"}, + {file = "torch-2.2.1+cpu-cp311-cp311-linux_x86_64.whl", hash = "sha256:6bc973d5632374b92b4b293817b4d2ff8c8ce1c784c748b471dba1fffcd9c333"}, + {file = "torch-2.2.1+cpu-cp311-cp311-win_amd64.whl", hash = "sha256:abdec34b0ade8fca0520055e72c3094425ae0ef210718e9c0278121cd3608c32"}, + {file = "torch-2.2.1+cpu-cp312-cp312-linux_x86_64.whl", hash = "sha256:d7339580135da4105c1244a8621faa076990409afeab5a7b642c3c1ee70a5622"}, + {file = "torch-2.2.1+cpu-cp312-cp312-win_amd64.whl", hash = "sha256:039128fcb5548122465b15f679b8831c47d14f0d6c28c1f1b631f8019c104720"}, + {file = "torch-2.2.1+cpu-cp38-cp38-linux_x86_64.whl", hash = "sha256:2b447f7bb50b393b4544b4036d587e39ab524d4353e77c197f6a2727f22b0d47"}, + {file = "torch-2.2.1+cpu-cp38-cp38-win_amd64.whl", hash = "sha256:2ccdf3e5f71e6426ea9e34d21c3cc333b29d4f48299b981d28aeb5112b5495e1"}, + {file = "torch-2.2.1+cpu-cp39-cp39-linux_x86_64.whl", hash = "sha256:2fb340b289760040a16a77a6d70b8a48961abba1822e6f58705c97c80befa03e"}, + {file = "torch-2.2.1+cpu-cp39-cp39-win_amd64.whl", hash = "sha256:e03dc4654ecceeb5b03f0a6f60b342c0e0d267b3ebc61e4f672cace1df8cd930"}, +] + +[package.dependencies] +filelock = "*" +fsspec = "*" +jinja2 = "*" +networkx = "*" +sympy = "*" +typing-extensions = ">=4.8.0" + +[package.extras] +opt-einsum = ["opt-einsum (>=3.3)"] +optree = ["optree (>=0.9.1)"] + +[package.source] +type = "legacy" +url = "https://download.pytorch.org/whl/cpu" +reference = "pytorch" + +[[package]] +name = "tqdm" +version = "4.66.2" +description = "Fast, Extensible Progress Meter" +optional = false +python-versions = ">=3.7" +files = [ + {file = "tqdm-4.66.2-py3-none-any.whl", hash = "sha256:1ee4f8a893eb9bef51c6e35730cebf234d5d0b6bd112b0271e10ed7c24a02bd9"}, + {file = "tqdm-4.66.2.tar.gz", hash = "sha256:6cd52cdf0fef0e0f543299cfc96fec90d7b8a7e88745f411ec33eb44d5ed3531"}, +] + +[package.dependencies] +colorama = {version = "*", markers = "platform_system == \"Windows\""} + +[package.extras] +dev = ["pytest (>=6)", "pytest-cov", "pytest-timeout", "pytest-xdist"] +notebook = ["ipywidgets (>=6)"] +slack = ["slack-sdk"] +telegram = ["requests"] + +[[package]] +name = "transformers" +version = "4.38.1" +description = "State-of-the-art Machine Learning for JAX, PyTorch and TensorFlow" +optional = false +python-versions = ">=3.8.0" +files = [ + {file = "transformers-4.38.1-py3-none-any.whl", hash = "sha256:a7a9265fb060183e9d975cbbadc4d531b10281589c43f6d07563f86322728973"}, + {file = "transformers-4.38.1.tar.gz", hash = "sha256:86dc84ccbe36123647e84cbd50fc31618c109a41e6be92514b064ab55bf1304c"}, +] + +[package.dependencies] +filelock = "*" +huggingface-hub = ">=0.19.3,<1.0" +numpy = ">=1.17" +packaging = ">=20.0" +pyyaml = ">=5.1" +regex = "!=2019.12.17" +requests = "*" +safetensors = ">=0.4.1" +tokenizers = ">=0.14,<0.19" +tqdm = ">=4.27" + +[package.extras] +accelerate = ["accelerate (>=0.21.0)"] +agents = ["Pillow (>=10.0.1,<=15.0)", "accelerate (>=0.21.0)", "datasets (!=2.5.0)", "diffusers", "opencv-python", "sentencepiece (>=0.1.91,!=0.1.92)", "torch"] +all = ["Pillow (>=10.0.1,<=15.0)", "accelerate (>=0.21.0)", "av (==9.2.0)", "codecarbon (==1.2.0)", "decord (==0.6.0)", "flax (>=0.4.1,<=0.7.0)", "jax (>=0.4.1,<=0.4.13)", "jaxlib (>=0.4.1,<=0.4.13)", "kenlm", "keras-nlp (>=0.3.1)", "librosa", "onnxconverter-common", "optax (>=0.0.8,<=0.1.4)", "optuna", "phonemizer", "protobuf", "pyctcdecode (>=0.4.0)", "ray[tune] (>=2.7.0)", "sentencepiece (>=0.1.91,!=0.1.92)", "sigopt", "tensorflow (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx", "timm", "tokenizers (>=0.14,<0.19)", "torch", "torchaudio", "torchvision"] +audio = ["kenlm", "librosa", "phonemizer", "pyctcdecode (>=0.4.0)"] +codecarbon = ["codecarbon (==1.2.0)"] +deepspeed = ["accelerate (>=0.21.0)", "deepspeed (>=0.9.3)"] +deepspeed-testing = ["GitPython (<3.1.19)", "accelerate (>=0.21.0)", "beautifulsoup4", "cookiecutter (==1.7.3)", "datasets (!=2.5.0)", "deepspeed (>=0.9.3)", "dill (<0.3.5)", "evaluate (>=0.2.0)", "faiss-cpu", "hf-doc-builder (>=0.3.0)", "nltk", "optuna", "parameterized", "protobuf", "psutil", "pydantic", "pytest (>=7.2.0,<8.0.0)", "pytest-timeout", "pytest-xdist", "rjieba", "rouge-score (!=0.0.7,!=0.0.8,!=0.1,!=0.1.1)", "ruff (==0.1.5)", "sacrebleu (>=1.4.12,<2.0.0)", "sacremoses", "sentencepiece (>=0.1.91,!=0.1.92)", "tensorboard", "timeout-decorator"] +dev = ["GitPython (<3.1.19)", "Pillow (>=10.0.1,<=15.0)", "accelerate (>=0.21.0)", "av (==9.2.0)", "beautifulsoup4", "codecarbon (==1.2.0)", "cookiecutter (==1.7.3)", "datasets (!=2.5.0)", "decord (==0.6.0)", "dill (<0.3.5)", "evaluate (>=0.2.0)", "faiss-cpu", "flax (>=0.4.1,<=0.7.0)", "fugashi (>=1.0)", "hf-doc-builder", "hf-doc-builder (>=0.3.0)", "ipadic (>=1.0.0,<2.0)", "isort (>=5.5.4)", "jax (>=0.4.1,<=0.4.13)", "jaxlib (>=0.4.1,<=0.4.13)", "kenlm", "keras-nlp (>=0.3.1)", "librosa", "nltk", "onnxconverter-common", "optax (>=0.0.8,<=0.1.4)", "optuna", "parameterized", "phonemizer", "protobuf", "psutil", "pyctcdecode (>=0.4.0)", "pydantic", "pytest (>=7.2.0,<8.0.0)", "pytest-timeout", "pytest-xdist", "ray[tune] (>=2.7.0)", "rhoknp (>=1.1.0,<1.3.1)", "rjieba", "rouge-score (!=0.0.7,!=0.0.8,!=0.1,!=0.1.1)", "ruff (==0.1.5)", "sacrebleu (>=1.4.12,<2.0.0)", "sacremoses", "scikit-learn", "sentencepiece (>=0.1.91,!=0.1.92)", "sigopt", "sudachidict-core (>=20220729)", "sudachipy (>=0.6.6)", "tensorboard", "tensorflow (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx", "timeout-decorator", "timm", "tokenizers (>=0.14,<0.19)", "torch", "torchaudio", "torchvision", "unidic (>=1.0.2)", "unidic-lite (>=1.0.7)", "urllib3 (<2.0.0)"] +dev-tensorflow = ["GitPython (<3.1.19)", "Pillow (>=10.0.1,<=15.0)", "beautifulsoup4", "cookiecutter (==1.7.3)", "datasets (!=2.5.0)", "dill (<0.3.5)", "evaluate (>=0.2.0)", "faiss-cpu", "hf-doc-builder", "hf-doc-builder (>=0.3.0)", "isort (>=5.5.4)", "kenlm", "keras-nlp (>=0.3.1)", "librosa", "nltk", "onnxconverter-common", "onnxruntime (>=1.4.0)", "onnxruntime-tools (>=1.4.2)", "parameterized", "phonemizer", "protobuf", "psutil", "pyctcdecode (>=0.4.0)", "pydantic", "pytest (>=7.2.0,<8.0.0)", "pytest-timeout", "pytest-xdist", "rjieba", "rouge-score (!=0.0.7,!=0.0.8,!=0.1,!=0.1.1)", "ruff (==0.1.5)", "sacrebleu (>=1.4.12,<2.0.0)", "sacremoses", "scikit-learn", "sentencepiece (>=0.1.91,!=0.1.92)", "tensorboard", "tensorflow (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx", "timeout-decorator", "tokenizers (>=0.14,<0.19)", "urllib3 (<2.0.0)"] +dev-torch = ["GitPython (<3.1.19)", "Pillow (>=10.0.1,<=15.0)", "accelerate (>=0.21.0)", "beautifulsoup4", "codecarbon (==1.2.0)", "cookiecutter (==1.7.3)", "datasets (!=2.5.0)", "dill (<0.3.5)", "evaluate (>=0.2.0)", "faiss-cpu", "fugashi (>=1.0)", "hf-doc-builder", "hf-doc-builder (>=0.3.0)", "ipadic (>=1.0.0,<2.0)", "isort (>=5.5.4)", "kenlm", "librosa", "nltk", "onnxruntime (>=1.4.0)", "onnxruntime-tools (>=1.4.2)", "optuna", "parameterized", "phonemizer", "protobuf", "psutil", "pyctcdecode (>=0.4.0)", "pydantic", "pytest (>=7.2.0,<8.0.0)", "pytest-timeout", "pytest-xdist", "ray[tune] (>=2.7.0)", "rhoknp (>=1.1.0,<1.3.1)", "rjieba", "rouge-score (!=0.0.7,!=0.0.8,!=0.1,!=0.1.1)", "ruff (==0.1.5)", "sacrebleu (>=1.4.12,<2.0.0)", "sacremoses", "scikit-learn", "sentencepiece (>=0.1.91,!=0.1.92)", "sigopt", "sudachidict-core (>=20220729)", "sudachipy (>=0.6.6)", "tensorboard", "timeout-decorator", "timm", "tokenizers (>=0.14,<0.19)", "torch", "torchaudio", "torchvision", "unidic (>=1.0.2)", "unidic-lite (>=1.0.7)", "urllib3 (<2.0.0)"] +docs = ["Pillow (>=10.0.1,<=15.0)", "accelerate (>=0.21.0)", "av (==9.2.0)", "codecarbon (==1.2.0)", "decord (==0.6.0)", "flax (>=0.4.1,<=0.7.0)", "hf-doc-builder", "jax (>=0.4.1,<=0.4.13)", "jaxlib (>=0.4.1,<=0.4.13)", "kenlm", "keras-nlp (>=0.3.1)", "librosa", "onnxconverter-common", "optax (>=0.0.8,<=0.1.4)", "optuna", "phonemizer", "protobuf", "pyctcdecode (>=0.4.0)", "ray[tune] (>=2.7.0)", "sentencepiece (>=0.1.91,!=0.1.92)", "sigopt", "tensorflow (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx", "timm", "tokenizers (>=0.14,<0.19)", "torch", "torchaudio", "torchvision"] +docs-specific = ["hf-doc-builder"] +flax = ["flax (>=0.4.1,<=0.7.0)", "jax (>=0.4.1,<=0.4.13)", "jaxlib (>=0.4.1,<=0.4.13)", "optax (>=0.0.8,<=0.1.4)"] +flax-speech = ["kenlm", "librosa", "phonemizer", "pyctcdecode (>=0.4.0)"] +ftfy = ["ftfy"] +integrations = ["optuna", "ray[tune] (>=2.7.0)", "sigopt"] +ja = ["fugashi (>=1.0)", "ipadic (>=1.0.0,<2.0)", "rhoknp (>=1.1.0,<1.3.1)", "sudachidict-core (>=20220729)", "sudachipy (>=0.6.6)", "unidic (>=1.0.2)", "unidic-lite (>=1.0.7)"] +modelcreation = ["cookiecutter (==1.7.3)"] +natten = ["natten (>=0.14.6,<0.15.0)"] +onnx = ["onnxconverter-common", "onnxruntime (>=1.4.0)", "onnxruntime-tools (>=1.4.2)", "tf2onnx"] +onnxruntime = ["onnxruntime (>=1.4.0)", "onnxruntime-tools (>=1.4.2)"] +optuna = ["optuna"] +quality = ["GitPython (<3.1.19)", "datasets (!=2.5.0)", "hf-doc-builder (>=0.3.0)", "isort (>=5.5.4)", "ruff (==0.1.5)", "urllib3 (<2.0.0)"] +ray = ["ray[tune] (>=2.7.0)"] +retrieval = ["datasets (!=2.5.0)", "faiss-cpu"] +sagemaker = ["sagemaker (>=2.31.0)"] +sentencepiece = ["protobuf", "sentencepiece (>=0.1.91,!=0.1.92)"] +serving = ["fastapi", "pydantic", "starlette", "uvicorn"] +sigopt = ["sigopt"] +sklearn = ["scikit-learn"] +speech = ["kenlm", "librosa", "phonemizer", "pyctcdecode (>=0.4.0)", "torchaudio"] +testing = ["GitPython (<3.1.19)", "beautifulsoup4", "cookiecutter (==1.7.3)", "datasets (!=2.5.0)", "dill (<0.3.5)", "evaluate (>=0.2.0)", "faiss-cpu", "hf-doc-builder (>=0.3.0)", "nltk", "parameterized", "protobuf", "psutil", "pydantic", "pytest (>=7.2.0,<8.0.0)", "pytest-timeout", "pytest-xdist", "rjieba", "rouge-score (!=0.0.7,!=0.0.8,!=0.1,!=0.1.1)", "ruff (==0.1.5)", "sacrebleu (>=1.4.12,<2.0.0)", "sacremoses", "tensorboard", "timeout-decorator"] +tf = ["keras-nlp (>=0.3.1)", "onnxconverter-common", "tensorflow (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx"] +tf-cpu = ["keras-nlp (>=0.3.1)", "onnxconverter-common", "tensorflow-cpu (>=2.6,<2.16)", "tensorflow-text (<2.16)", "tf2onnx"] +tf-speech = ["kenlm", "librosa", "phonemizer", "pyctcdecode (>=0.4.0)"] +timm = ["timm"] +tokenizers = ["tokenizers (>=0.14,<0.19)"] +torch = ["accelerate (>=0.21.0)", "torch"] +torch-speech = ["kenlm", "librosa", "phonemizer", "pyctcdecode (>=0.4.0)", "torchaudio"] +torch-vision = ["Pillow (>=10.0.1,<=15.0)", "torchvision"] +torchhub = ["filelock", "huggingface-hub (>=0.19.3,<1.0)", "importlib-metadata", "numpy (>=1.17)", "packaging (>=20.0)", "protobuf", "regex (!=2019.12.17)", "requests", "sentencepiece (>=0.1.91,!=0.1.92)", "tokenizers (>=0.14,<0.19)", "torch", "tqdm (>=4.27)"] +video = ["av (==9.2.0)", "decord (==0.6.0)"] +vision = ["Pillow (>=10.0.1,<=15.0)"] + +[[package]] +name = "typing-extensions" +version = "4.9.0" +description = "Backported and Experimental Type Hints for Python 3.8+" +optional = false +python-versions = ">=3.8" +files = [ + {file = "typing_extensions-4.9.0-py3-none-any.whl", hash = "sha256:af72aea155e91adfc61c3ae9e0e342dbc0cba726d6cba4b6c72c1f34e47291cd"}, + {file = "typing_extensions-4.9.0.tar.gz", hash = "sha256:23478f88c37f27d76ac8aee6c905017a143b0b1b886c3c9f66bc2fd94f9f5783"}, +] + +[[package]] +name = "urllib3" +version = "2.2.1" +description = "HTTP library with thread-safe connection pooling, file post, and more." +optional = false +python-versions = ">=3.8" +files = [ + {file = "urllib3-2.2.1-py3-none-any.whl", hash = "sha256:450b20ec296a467077128bff42b73080516e71b56ff59a60a02bef2232c4fa9d"}, + {file = "urllib3-2.2.1.tar.gz", hash = "sha256:d0570876c61ab9e520d776c38acbbb5b05a776d3f9ff98a5c8fd5162a444cf19"}, +] + +[package.extras] +brotli = ["brotli (>=1.0.9)", "brotlicffi (>=0.8.0)"] +h2 = ["h2 (>=4,<5)"] +socks = ["pysocks (>=1.5.6,!=1.5.7,<2.0)"] +zstd = ["zstandard (>=0.18.0)"] + +[[package]] +name = "wcwidth" +version = "0.2.13" +description = "Measures the displayed width of unicode strings in a terminal" +optional = false +python-versions = "*" +files = [ + {file = "wcwidth-0.2.13-py2.py3-none-any.whl", hash = "sha256:3da69048e4540d84af32131829ff948f1e022c1c6bdb8d6102117aac784f6859"}, + {file = "wcwidth-0.2.13.tar.gz", hash = "sha256:72ea0c06399eb286d978fdedb6923a9eb47e1c486ce63e9b4e64fc18303972b5"}, +] + +[metadata] +lock-version = "2.0" +python-versions = ">=3.9" +content-hash = "c8c4cc87637266a7b85debcbafa8887c5ad81cc8ef40e98a3f52c7c50af05c03" diff --git a/pyproject.toml b/pyproject.toml new file mode 100644 index 000000000..25e2e20b2 --- /dev/null +++ b/pyproject.toml @@ -0,0 +1,44 @@ +[tool.poetry] +name = "llama-cpp-scripts" +version = "0.0.0" +description = "Scripts that ship with llama.cpp" +authors = ["GGML "] +readme = "README.md" +homepage = "https://ggml.ai" +repository = "https://github.com/ggerganov/llama.cpp" +keywords = ["ggml", "gguf", "llama.cpp"] +packages = [{ include = "*.py", from = "." }] +classifiers = [ + "Programming Language :: Python :: 3", + "License :: OSI Approved :: MIT License", + "Operating System :: OS Independent", +] + +[tool.poetry.dependencies] +python = ">=3.9" +numpy = "^1.25.0" +sentencepiece = ">=0.1.98,<0.2.0" +transformers = ">=4.35.2,<5.0.0" +protobuf = ">=4.21.0,<5.0.0" +gguf = { path = "./gguf-py" } +torch = { version = "^2.2.0", source = "pytorch" } + +[tool.poetry.dev-dependencies] +pytest = "^5.2" + + +# Force wheel + cpu +# For discussion and context see https://github.com/python-poetry/poetry#6409 +[[tool.poetry.source]] +name = "pytorch" +url = "https://download.pytorch.org/whl/cpu" +priority = "explicit" + +[build-system] +requires = ["poetry-core>=1.0.0"] +build-backend = "poetry.core.masonry.api" + +[tool.poetry.scripts] +llama-convert-hf-to-gguf = "convert_hf_to_gguf:main" +llama-convert-llama-ggml-to-gguf = "convert_llama_ggml_to_gguf:main" +llama-ggml-vk-generate-shaders = "ggml_vk_generate_shaders:main" diff --git a/requirements.txt b/requirements.txt index e5cfbf10b..52456c2e6 100644 --- a/requirements.txt +++ b/requirements.txt @@ -4,8 +4,8 @@ # Package versions must stay compatible across all top-level python scripts. # --r ./requirements/requirements-convert-legacy-llama.txt +-r ./requirements/requirements-convert_legacy_llama.txt --r ./requirements/requirements-convert-hf-to-gguf.txt --r ./requirements/requirements-convert-hf-to-gguf-update.txt --r ./requirements/requirements-convert-llama-ggml-to-gguf.txt +-r ./requirements/requirements-convert_hf_to_gguf.txt +-r ./requirements/requirements-convert_hf_to_gguf_update.txt +-r ./requirements/requirements-convert_llama_ggml_to_gguf.txt diff --git a/requirements/requirements-convert-hf-to-gguf-update.txt b/requirements/requirements-convert-hf-to-gguf-update.txt deleted file mode 100644 index a7112f396..000000000 --- a/requirements/requirements-convert-hf-to-gguf-update.txt +++ /dev/null @@ -1,2 +0,0 @@ --r ./requirements-convert-legacy-llama.txt -torch~=2.2.1 diff --git a/requirements/requirements-convert-hf-to-gguf.txt b/requirements/requirements-convert-hf-to-gguf.txt deleted file mode 100644 index a7112f396..000000000 --- a/requirements/requirements-convert-hf-to-gguf.txt +++ /dev/null @@ -1,2 +0,0 @@ --r ./requirements-convert-legacy-llama.txt -torch~=2.2.1 diff --git a/requirements/requirements-convert-llama-ggml-to-gguf.txt b/requirements/requirements-convert-llama-ggml-to-gguf.txt deleted file mode 100644 index e80c29012..000000000 --- a/requirements/requirements-convert-llama-ggml-to-gguf.txt +++ /dev/null @@ -1 +0,0 @@ --r ./requirements-convert-legacy-llama.txt diff --git a/requirements/requirements-convert_hf_to_gguf.txt b/requirements/requirements-convert_hf_to_gguf.txt new file mode 100644 index 000000000..653355c07 --- /dev/null +++ b/requirements/requirements-convert_hf_to_gguf.txt @@ -0,0 +1,2 @@ +-r ./requirements-convert_legacy_llama.txt +torch~=2.2.1 diff --git a/requirements/requirements-convert_hf_to_gguf_update.txt b/requirements/requirements-convert_hf_to_gguf_update.txt new file mode 100644 index 000000000..653355c07 --- /dev/null +++ b/requirements/requirements-convert_hf_to_gguf_update.txt @@ -0,0 +1,2 @@ +-r ./requirements-convert_legacy_llama.txt +torch~=2.2.1 diff --git a/requirements/requirements-convert-legacy-llama.txt b/requirements/requirements-convert_legacy_llama.txt similarity index 100% rename from requirements/requirements-convert-legacy-llama.txt rename to requirements/requirements-convert_legacy_llama.txt diff --git a/requirements/requirements-convert_llama_ggml_to_gguf.txt b/requirements/requirements-convert_llama_ggml_to_gguf.txt new file mode 100644 index 000000000..afe2747d4 --- /dev/null +++ b/requirements/requirements-convert_llama_ggml_to_gguf.txt @@ -0,0 +1 @@ +-r ./requirements-convert_legacy_llama.txt diff --git a/scripts/check-requirements.sh b/scripts/check-requirements.sh index 0c6afdd59..48f924c02 100755 --- a/scripts/check-requirements.sh +++ b/scripts/check-requirements.sh @@ -97,9 +97,9 @@ check_requirements() { } check_convert_script() { - local py=$1 # e.g. ./convert-hf-to-gguf.py - local pyname=${py##*/} # e.g. convert-hf-to-gguf.py - pyname=${pyname%.py} # e.g. convert-hf-to-gguf + local py=$1 # e.g. ./convert_hf_to_gguf.py + local pyname=${py##*/} # e.g. convert_hf_to_gguf.py + pyname=${pyname%.py} # e.g. convert_hf_to_gguf info "$py: beginning check" @@ -166,12 +166,12 @@ if (( do_cleanup )); then rm -rf -- "$all_venv" fi -check_convert_script examples/convert-legacy-llama.py -for py in convert-*.py; do - # skip convert-hf-to-gguf-update.py +check_convert_script examples/convert_legacy_llama.py +for py in convert_*.py; do + # skip convert_hf_to_gguf_update.py # TODO: the check is failing for some reason: # https://github.com/ggerganov/llama.cpp/actions/runs/8875330981/job/24364557177?pr=6920 - [[ $py == convert-hf-to-gguf-update.py ]] && continue + [[ $py == convert_hf_to_gguf_update.py ]] && continue check_convert_script "$py" done diff --git a/scripts/convert-gg.sh b/scripts/convert-gg.sh deleted file mode 100755 index 8a0168432..000000000 --- a/scripts/convert-gg.sh +++ /dev/null @@ -1,26 +0,0 @@ -#!/bin/bash - -set -e - -# LLaMA v1 -python3 examples/convert-legacy-llama.py ../llama1/7B --outfile models/llama-7b/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../llama1/13B --outfile models/llama-13b/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../llama1/30B --outfile models/llama-30b/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../llama1/65B --outfile models/llama-65b/ggml-model-f16.gguf --outtype f16 - -# LLaMA v2 -python3 examples/convert-legacy-llama.py ../llama2/llama-2-7b --outfile models/llama-7b-v2/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../llama2/llama-2-13b --outfile models/llama-13b-v2/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../llama2/llama-2-70b --outfile models/llama-70b-v2/ggml-model-f16.gguf --outtype f16 - -# Code Llama -python3 examples/convert-legacy-llama.py ../codellama/CodeLlama-7b/ --outfile models/codellama-7b/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../codellama/CodeLlama-13b/ --outfile models/codellama-13b/ggml-model-f16.gguf --outtype f16 -python3 examples/convert-legacy-llama.py ../codellama/CodeLlama-34b/ --outfile models/codellama-34b/ggml-model-f16.gguf --outtype f16 - -# Falcon -python3 convert-falcon-hf-to-gguf.py ../falcon/falcon-7b 1 -mv -v ../falcon/falcon-7b/ggml-model-f16.gguf models/falcon-7b/ggml-model-f16.gguf - -python3 convert-falcon-hf-to-gguf.py ../falcon/falcon-40b 1 -mv -v ../falcon/falcon-40b/ggml-model-f16.gguf models/falcon-40b/ggml-model-f16.gguf diff --git a/scripts/pod-llama.sh b/scripts/pod-llama.sh index 0d6d4032d..6e56e1ed0 100644 --- a/scripts/pod-llama.sh +++ b/scripts/pod-llama.sh @@ -75,7 +75,7 @@ if [ "$1" -eq "1" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/tinyllama-1b --outfile ./models/tinyllama-1b/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/tinyllama-1b --outfile ./models/tinyllama-1b/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/tinyllama-1b/ggml-model-f16.gguf ./models/tinyllama-1b/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/tinyllama-1b/ggml-model-f16.gguf ./models/tinyllama-1b/ggml-model-q4_k.gguf q4_k @@ -90,7 +90,7 @@ if [ "$1" -eq "2" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-7b --outfile ./models/codellama-7b/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-7b --outfile ./models/codellama-7b/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-7b/ggml-model-f16.gguf ./models/codellama-7b/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-7b/ggml-model-f16.gguf ./models/codellama-7b/ggml-model-q4_k.gguf q4_k @@ -105,7 +105,7 @@ if [ "$1" -eq "3" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-13b --outfile ./models/codellama-13b/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-13b --outfile ./models/codellama-13b/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-13b/ggml-model-f16.gguf ./models/codellama-13b/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-13b/ggml-model-f16.gguf ./models/codellama-13b/ggml-model-q4_k.gguf q4_k @@ -120,7 +120,7 @@ if [ "$1" -eq "4" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-34b --outfile ./models/codellama-34b/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-34b --outfile ./models/codellama-34b/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-34b/ggml-model-f16.gguf ./models/codellama-34b/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-34b/ggml-model-f16.gguf ./models/codellama-34b/ggml-model-q4_k.gguf q4_k @@ -135,7 +135,7 @@ if [ "$1" -eq "5" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-7b-instruct --outfile ./models/codellama-7b-instruct/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-7b-instruct --outfile ./models/codellama-7b-instruct/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-7b-instruct/ggml-model-f16.gguf ./models/codellama-7b-instruct/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-7b-instruct/ggml-model-f16.gguf ./models/codellama-7b-instruct/ggml-model-q4_k.gguf q4_k @@ -150,7 +150,7 @@ if [ "$1" -eq "6" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-13b-instruct --outfile ./models/codellama-13b-instruct/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-13b-instruct --outfile ./models/codellama-13b-instruct/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-13b-instruct/ggml-model-f16.gguf ./models/codellama-13b-instruct/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-13b-instruct/ggml-model-f16.gguf ./models/codellama-13b-instruct/ggml-model-q4_k.gguf q4_k @@ -165,7 +165,7 @@ if [ "$1" -eq "7" ]; then cd /workspace/llama.cpp - python3 examples/convert-legacy-llama.py ./models/codellama-34b-instruct --outfile ./models/codellama-34b-instruct/ggml-model-f16.gguf --outtype f16 + python3 examples/convert_legacy_llama.py ./models/codellama-34b-instruct --outfile ./models/codellama-34b-instruct/ggml-model-f16.gguf --outtype f16 ./llama-quantize ./models/codellama-34b-instruct/ggml-model-f16.gguf ./models/codellama-34b-instruct/ggml-model-q4_0.gguf q4_0 ./llama-quantize ./models/codellama-34b-instruct/ggml-model-f16.gguf ./models/codellama-34b-instruct/ggml-model-q4_k.gguf q4_k diff --git a/src/CMakeLists.txt b/src/CMakeLists.txt index ccb607e56..c2049df79 100644 --- a/src/CMakeLists.txt +++ b/src/CMakeLists.txt @@ -1,7 +1,5 @@ # TODO: should not use this if (WIN32) - add_compile_definitions(_CRT_SECURE_NO_WARNINGS) - if (BUILD_SHARED_LIBS) set(CMAKE_WINDOWS_EXPORT_ALL_SYMBOLS ON) endif() diff --git a/src/llama.cpp b/src/llama.cpp index eb1bde269..824540777 100644 --- a/src/llama.cpp +++ b/src/llama.cpp @@ -104,8 +104,10 @@ #define LLAMA_ATTRIBUTE_FORMAT(...) #endif +// bump if necessary #define LLAMA_MAX_NODES 8192 -#define LLAMA_MAX_EXPERTS 160 +#define LLAMA_MAX_LAYERS 256 +#define LLAMA_MAX_EXPERTS 160 // DeepSeekV2 // // logging @@ -224,6 +226,7 @@ enum llm_arch { LLM_ARCH_COMMAND_R, LLM_ARCH_DBRX, LLM_ARCH_OLMO, + LLM_ARCH_OPENELM, LLM_ARCH_ARCTIC, LLM_ARCH_DEEPSEEK2, LLM_ARCH_CHATGLM, @@ -267,6 +270,7 @@ static const std::map LLM_ARCH_NAMES = { { LLM_ARCH_COMMAND_R, "command-r" }, { LLM_ARCH_DBRX, "dbrx" }, { LLM_ARCH_OLMO, "olmo" }, + { LLM_ARCH_OPENELM, "openelm" }, { LLM_ARCH_ARCTIC, "arctic" }, { LLM_ARCH_DEEPSEEK2, "deepseek2" }, { LLM_ARCH_CHATGLM, "chatglm" }, @@ -1136,6 +1140,22 @@ static const std::map> LLM_TENSOR_NA { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, }, }, + { + LLM_ARCH_OPENELM, + { + { LLM_TENSOR_TOKEN_EMBD, "token_embd" }, + { LLM_TENSOR_OUTPUT_NORM, "output_norm" }, + { LLM_TENSOR_ATTN_NORM, "blk.%d.attn_norm" }, + { LLM_TENSOR_ATTN_QKV, "blk.%d.attn_qkv" }, + { LLM_TENSOR_ATTN_Q_NORM, "blk.%d.attn_q_norm" }, + { LLM_TENSOR_ATTN_K_NORM, "blk.%d.attn_k_norm" }, + { LLM_TENSOR_ATTN_OUT, "blk.%d.attn_output" }, + { LLM_TENSOR_FFN_NORM, "blk.%d.ffn_norm" }, + { LLM_TENSOR_FFN_GATE, "blk.%d.ffn_gate" }, + { LLM_TENSOR_FFN_DOWN, "blk.%d.ffn_down" }, + { LLM_TENSOR_FFN_UP, "blk.%d.ffn_up" }, + }, + }, { LLM_ARCH_ARCTIC, { @@ -1992,18 +2012,19 @@ using llama_mlocks = std::vector>; // NOTE: avoid ever using this except for building the token_to_piece caches static std::string llama_token_to_piece(const struct llama_model * model, llama_token token, bool special) { - std::vector result(8, 0); - const int n_tokens = llama_token_to_piece(model, token, result.data(), result.size(), special); - if (n_tokens < 0) { - result.resize(-n_tokens); - int check = llama_token_to_piece(model, token, result.data(), result.size(), special); - GGML_ASSERT(check == -n_tokens); + std::string piece; + piece.resize(piece.capacity()); // using string internal cache + const int n_chars = llama_token_to_piece(model, token, &piece[0], piece.size(), 0, special); + if (n_chars < 0) { + piece.resize(-n_chars); + int check = llama_token_to_piece(model, token, &piece[0], piece.size(), 0, special); + GGML_ASSERT(check == -n_chars); } else { - result.resize(n_tokens); + piece.resize(n_chars); } - return std::string(result.data(), result.size()); + return piece; } static ggml_backend_buffer_type_t llama_default_buffer_type_cpu(bool host_buffer) { @@ -2061,12 +2082,20 @@ enum e_model { MODEL_17M, MODEL_22M, MODEL_33M, + MODEL_60M, MODEL_70M, + MODEL_80M, MODEL_109M, MODEL_137M, MODEL_160M, + MODEL_220M, + MODEL_250M, + MODEL_270M, MODEL_335M, MODEL_410M, + MODEL_450M, + MODEL_770M, + MODEL_780M, MODEL_0_5B, MODEL_1B, MODEL_1_3B, @@ -2080,6 +2109,7 @@ enum e_model { MODEL_7B, MODEL_8B, MODEL_9B, + MODEL_11B, MODEL_12B, MODEL_13B, MODEL_14B, @@ -2119,17 +2149,19 @@ struct llama_hparams { uint32_t n_vocab; uint32_t n_ctx_train; // context size the model was trained on uint32_t n_embd; - uint32_t n_head; - uint32_t n_head_kv; uint32_t n_layer; uint32_t n_rot; uint32_t n_swa = 0; // sliding window attention (SWA) uint32_t n_embd_head_k; // dimension of keys (d_k). d_q is assumed to be the same, but there are n_head q heads, and only n_head_kv k-v heads uint32_t n_embd_head_v; // dimension of values (d_v) aka n_embd_head - uint32_t n_ff; uint32_t n_expert = 0; uint32_t n_expert_used = 0; uint32_t n_vocab_type = 0; // for BERT-style token types + uint32_t n_rel_attn_bkts = 0; + + std::array n_head_arr; + std::array n_head_kv_arr; + std::array n_ff_arr; uint32_t n_layer_dense_lead = 0; uint32_t n_lora_q = 0; @@ -2165,6 +2197,10 @@ struct llama_hparams { bool use_alibi = false; bool attn_soft_cap = false; + // needed by encoder-decoder models (e.g. T5, FLAN-T5) + // ref: https://github.com/ggerganov/llama.cpp/pull/8141 + llama_token dec_start_token_id = -1; + enum llama_pooling_type pooling_type = LLAMA_POOLING_TYPE_NONE; enum llama_rope_type rope_type = LLAMA_ROPE_TYPE_NONE; enum llama_rope_scaling_type rope_scaling_type_train = LLAMA_ROPE_SCALING_TYPE_NONE; @@ -2174,17 +2210,19 @@ struct llama_hparams { if (this->n_vocab != other.n_vocab) return true; if (this->n_ctx_train != other.n_ctx_train) return true; if (this->n_embd != other.n_embd) return true; - if (this->n_head != other.n_head) return true; - if (this->n_head_kv != other.n_head_kv) return true; if (this->n_layer != other.n_layer) return true; if (this->n_rot != other.n_rot) return true; if (this->n_swa != other.n_swa) return true; if (this->n_embd_head_k != other.n_embd_head_k) return true; if (this->n_embd_head_v != other.n_embd_head_v) return true; - if (this->n_ff != other.n_ff) return true; if (this->n_expert != other.n_expert) return true; if (this->n_expert_used != other.n_expert_used) return true; + if (this->n_head_arr != other.n_head_arr) return true; + if (this->n_head_kv_arr != other.n_head_kv_arr) return true; + if (this->n_ff_arr != other.n_ff_arr) return true; + + if (this->n_rel_attn_bkts != other.n_rel_attn_bkts) return true; if (this->n_layer_dense_lead != other.n_layer_dense_lead) return true; if (this->n_lora_q != other.n_lora_q) return true; if (this->n_lora_kv != other.n_lora_kv) return true; @@ -2200,6 +2238,8 @@ struct llama_hparams { if (this->ssm_d_state != other.ssm_d_state) return true; if (this->ssm_dt_rank != other.ssm_dt_rank) return true; + if (this->dec_start_token_id != other.dec_start_token_id) return true; + const float EPSILON = 1e-9f; if (!is_float_close(this->f_norm_eps, other.f_norm_eps, EPSILON)) return true; @@ -2213,18 +2253,53 @@ struct llama_hparams { return false; } - uint32_t n_gqa() const { + uint32_t n_head(uint32_t il = 0) const { + if (il < n_layer) { + return n_head_arr[il]; + } + + GGML_ASSERT(false); + return 0; + } + + uint32_t n_head_kv(uint32_t il = 0) const { + if (il < n_layer) { + return n_head_kv_arr[il]; + } + + GGML_ASSERT(false); + return 0; + } + + uint32_t n_ff(uint32_t il = 0) const { + if (il < n_layer) { + return n_ff_arr[il]; + } + + GGML_ASSERT(false); + return 0; + } + + uint32_t n_gqa(uint32_t il = 0) const { + const uint32_t n_head = this->n_head(il); + const uint32_t n_head_kv = this->n_head_kv(il); + if (n_head_kv == 0) { return 0; } + return n_head/n_head_kv; } - uint32_t n_embd_k_gqa() const { // dimension of key embeddings across all k-v heads + uint32_t n_embd_k_gqa(uint32_t il = 0) const { // dimension of key embeddings across all k-v heads + const uint32_t n_head_kv = this->n_head_kv(il); + return n_embd_head_k * n_head_kv; } - uint32_t n_embd_v_gqa() const { // dimension of value embeddings across all k-v heads + uint32_t n_embd_v_gqa(uint32_t il = 0) const { // dimension of value embeddings across all k-v heads + const uint32_t n_head_kv = this->n_head_kv(il); + return n_embd_head_v * n_head_kv; } @@ -2241,6 +2316,8 @@ struct llama_hparams { } }; +static_assert(std::is_trivially_copyable::value, "llama_hparams must be trivially copyable"); + struct llama_cparams { uint32_t n_ctx; // context size used during inference uint32_t n_batch; @@ -2272,6 +2349,7 @@ struct llama_cparams { void * cb_eval_user_data; }; +// TODO: separate into "llama_layer_enc" and "llama_layer_dec" struct llama_layer { // normalization struct ggml_tensor * attn_norm; @@ -2289,6 +2367,8 @@ struct llama_layer { struct ggml_tensor * attn_sub_norm; struct ggml_tensor * attn_post_norm; struct ggml_tensor * ffn_sub_norm; + struct ggml_tensor * attn_norm_cross; + struct ggml_tensor * attn_norm_enc; // attention struct ggml_tensor * wq; @@ -2300,6 +2380,14 @@ struct llama_layer { struct ggml_tensor * wq_b; struct ggml_tensor * wkv_a_mqa; struct ggml_tensor * wkv_b; + struct ggml_tensor * wq_cross; + struct ggml_tensor * wk_cross; + struct ggml_tensor * wv_cross; + struct ggml_tensor * wo_cross; + struct ggml_tensor * wq_enc; + struct ggml_tensor * wk_enc; + struct ggml_tensor * wv_enc; + struct ggml_tensor * wo_enc; // attention bias struct ggml_tensor * bq; @@ -2308,6 +2396,11 @@ struct llama_layer { struct ggml_tensor * bo; struct ggml_tensor * bqkv; + // relative position bias + struct ggml_tensor * attn_rel_b; + struct ggml_tensor * attn_rel_b_enc; + struct ggml_tensor * attn_rel_b_cross; + // normalization struct ggml_tensor * ffn_norm; struct ggml_tensor * ffn_norm_b; @@ -2315,11 +2408,15 @@ struct llama_layer { struct ggml_tensor * layer_out_norm; struct ggml_tensor * layer_out_norm_b; struct ggml_tensor * ffn_norm_exps; + struct ggml_tensor * ffn_norm_enc; // ff struct ggml_tensor * ffn_gate; // w1 struct ggml_tensor * ffn_down; // w2 struct ggml_tensor * ffn_up; // w3 + struct ggml_tensor * ffn_gate_enc; + struct ggml_tensor * ffn_down_enc; + struct ggml_tensor * ffn_up_enc; // ff MoE struct ggml_tensor * ffn_gate_inp; @@ -2508,10 +2605,11 @@ struct llama_vocab { id special_eot_id = -1; // TODO: move above after "eos_id", and here add "file separator" token // tokenizer flags - bool tokenizer_add_space_prefix = true; + bool tokenizer_add_space_prefix = false; bool tokenizer_add_bos = false; bool tokenizer_add_eos = false; bool tokenizer_ignore_merges = false; + bool tokenizer_clean_spaces = false; // clean_up_tokenization_spaces bool tokenizer_remove_extra_whitespaces = false; bool tokenizer_escape_whitespaces = true; bool tokenizer_treat_whitespace_as_suffix = false; @@ -2553,6 +2651,7 @@ struct llama_model { struct ggml_tensor * output_norm_b; struct ggml_tensor * output; struct ggml_tensor * output_b; + struct ggml_tensor * output_norm_enc; std::vector layers; @@ -2681,6 +2780,13 @@ struct llama_context { // populated only when pooling_type != LLAMA_POOLING_TYPE_NONE std::map> embd_seq; + // whether we are computing encoder output or decoder output + bool is_encoding = false; + + // output of the encoder part of the encoder-decoder models + std::vector embd_enc; + std::vector> seq_ids_enc; + // memory buffers used to evaluate the model std::vector buf_compute_meta; ggml_backend_sched_t sched = nullptr; @@ -2701,6 +2807,9 @@ struct llama_context { struct ggml_tensor * inp_s_copy; // I32 [kv_size] struct ggml_tensor * inp_s_mask; // F32 [1, n_kv] struct ggml_tensor * inp_s_seq; // I32 [n_kv, n_batch] + struct ggml_tensor * inp_pos_bucket; // I32 [n_batch|n_kv, n_batch] + struct ggml_tensor * inp_embd_enc; // F32 [n_embd, n_outputs_enc] + struct ggml_tensor * inp_KQ_mask_cross; // F32 [n_outputs_enc, n_batch] // control vectors struct llama_control_vector cvec; @@ -2829,9 +2938,7 @@ static bool llama_kv_cache_init( const struct llama_hparams & hparams = model.hparams; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); - const int64_t n_layer = hparams.n_layer; + const int64_t n_layer = hparams.n_layer; cache.has_shift = false; @@ -2839,13 +2946,6 @@ static bool llama_kv_cache_init( cache.recurrent = model.arch == LLM_ARCH_MAMBA; cache.v_trans = !cparams.flash_attn; - // TODO: support mixed recurrent Transformer architectures - // NOTE: (!a || b) is a logical implication (a -> b) - GGML_ASSERT(!cache.recurrent || n_embd_k_gqa == hparams.n_embd_k_s()); - GGML_ASSERT(!cache.recurrent || n_embd_v_gqa == hparams.n_embd_v_s()); - GGML_ASSERT( cache.recurrent || n_embd_k_gqa == hparams.n_embd_k_gqa()); - GGML_ASSERT( cache.recurrent || n_embd_v_gqa == hparams.n_embd_v_gqa()); - cache.head = 0; cache.size = kv_size; cache.used = 0; @@ -2895,6 +2995,9 @@ static bool llama_kv_cache_init( cache.v_l.reserve(n_layer); for (int i = 0; i < (int) n_layer; i++) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(i) + hparams.n_embd_k_s(); + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(i) + hparams.n_embd_v_s(); + struct ggml_context * ctx = offload ? ctx_map.at(model.buft_layer[i].buft) : cache.ctxs.front(); ggml_tensor * k = ggml_new_tensor_1d(ctx, type_k, n_embd_k_gqa*kv_size); ggml_tensor * v = ggml_new_tensor_1d(ctx, type_v, n_embd_v_gqa*kv_size); @@ -3175,6 +3278,8 @@ static void llama_kv_cache_seq_add( if (p0 < 0) p0 = 0; if (p1 < 0) p1 = std::numeric_limits::max(); + // If there is no range then return early to avoid looping over the cache. + if (p0 == p1) return; if (cache.recurrent) { // for Mamba-like models, only the pos needs to be shifted @@ -3219,6 +3324,8 @@ static void llama_kv_cache_seq_div( int d) { if (p0 < 0) p0 = 0; if (p1 < 0) p1 = std::numeric_limits::max(); + // If there is no range then return early to avoid looping over the cache. + if (p0 == p1) return; if (cache.recurrent) { // for Mamba-like models, only the pos needs to be changed @@ -3770,9 +3877,9 @@ struct llama_model_loader { bool get_arr(const std::string & key, std::vector & result, const bool required = true) { const int kid = gguf_find_key(meta, key.c_str()); - if (kid < 0) { + if (kid < 0 || gguf_get_kv_type(meta, kid) != GGUF_TYPE_ARRAY) { if (required) { - throw std::runtime_error(format("key not found in model: %s", key.c_str())); + throw std::runtime_error(format("array key not found in model: %s", key.c_str())); } return false; } @@ -3780,22 +3887,53 @@ struct llama_model_loader { struct GGUFMeta::ArrayInfo arr_info = GGUFMeta::GKV::get_kv(meta, kid); - if (arr_info.gt != GGUF_TYPE_FLOAT32 && arr_info.gt != GGUF_TYPE_INT32) { - throw std::runtime_error(format("%s is not a float32 or int32 array", key.c_str())); + switch (arr_info.gt) { + case GGUF_TYPE_FLOAT32: GGML_ASSERT((std::is_same::value)); break; + case GGUF_TYPE_INT32: GGML_ASSERT( + (std::is_same::value) || + (std::is_same::value)); break; + default: + throw std::runtime_error(format("%s is not a float32, int32 array", key.c_str())); } - // GGML_ASSERT(gguf_type_size(arr_info.gt) == sizeof(T)); - GGML_ASSERT((arr_info.gt != GGUF_TYPE_FLOAT32 || std::is_same::value)); - GGML_ASSERT((arr_info.gt != GGUF_TYPE_INT32 || std::is_same::value)); - result.resize(arr_info.length); result.assign((const T*)arr_info.data, (const T *)arr_info.data + arr_info.length); return true; } + template + bool get_arr(const std::string & key, std::array & result, const bool required = true) { + const int kid = gguf_find_key(meta, key.c_str()); + + if (kid < 0 || gguf_get_kv_type(meta, kid) != GGUF_TYPE_ARRAY) { + if (required) { + throw std::runtime_error(format("array key not found in model: %s", key.c_str())); + } + return false; + } + + struct GGUFMeta::ArrayInfo arr_info = + GGUFMeta::GKV::get_kv(meta, kid); + + switch (arr_info.gt) { + case GGUF_TYPE_FLOAT32: GGML_ASSERT((std::is_same::value)); break; + case GGUF_TYPE_INT32: GGML_ASSERT( + (std::is_same::value) || + (std::is_same::value)); break; + default: + throw std::runtime_error(format("%s is not a float32, int32 array", key.c_str())); + } + + GGML_ASSERT(arr_info.length <= N_MAX); + + std::copy((const T*)arr_info.data, (const T *)arr_info.data + arr_info.length, result.begin()); + + return true; + } + template - bool get_arr(const enum llm_kv kid, T& result, const bool required = true) { + bool get_arr(const enum llm_kv kid, T & result, const bool required = true) { return get_arr(llm_kv(kid), result, required); } @@ -3820,6 +3958,50 @@ struct llama_model_loader { return get_key(llm_kv(kid), result, required); } + // get array of n <= N_MAX elements, or a single element repeated n times + template + bool get_key_or_arr(const std::string & key, std::array & result, uint32_t n, const bool required = true) { + GGML_ASSERT(n <= N_MAX); + + const int kid = gguf_find_key(meta, key.c_str()); + + if (kid < 0) { + if (required) { + throw std::runtime_error(format("key not found in model: %s", key.c_str())); + } + return false; + } + + if (gguf_get_kv_type(meta, kid) == GGUF_TYPE_ARRAY) { + struct GGUFMeta::ArrayInfo arr_info = + GGUFMeta::GKV::get_kv(meta, kid); + + if (n != arr_info.length) { + throw std::runtime_error(format("key %s has wrong array length; expected %u, got %u", key.c_str(), n, (uint32_t) arr_info.length)); + } + + return get_arr(key, result, required); + } else { + T value; + + bool ok = get_key(key, value, required); + if (!ok) { + return false; + } + + for (uint32_t i = 0; i < n; i++) { + result[i] = value; + } + + return true; + } + } + + template + bool get_key_or_arr(const enum llm_kv kid, T & result, uint32_t n, const bool required = true) { + return get_key_or_arr(llm_kv(kid), result, n, required); + } + std::string get_arch_name() const { return arch_name; } @@ -4052,7 +4234,7 @@ struct llama_model_loader { #if defined(GGML_USE_CUDA) // 4 staging buffers for async uploads, each sized 1MB seems to be a good default for single NVMe drives. // NVMe raid configurations might require more / larger buffers. - constexpr size_t num_buffers = 4; + constexpr size_t n_buffers = 4; constexpr size_t buffer_size = 1 * 1024 * 1024; // 1MB std::vector host_buffers; @@ -4078,7 +4260,7 @@ struct llama_model_loader { // If the cuda backend is active create pinned memory buffers and events for synchronisation. if (cuda_backend) { - for (size_t idx = 0; idx < num_buffers; ++idx) { + for (size_t idx = 0; idx < n_buffers; ++idx) { host_buffers.emplace_back(ggml_backend_buft_alloc_buffer(llama_default_buffer_type_cpu(true), buffer_size)); host_ptrs.emplace_back(ggml_backend_buffer_get_base(host_buffers[idx])); events.emplace_back(ggml_backend_event_new(cuda_backend)); @@ -4159,7 +4341,7 @@ struct llama_model_loader { bytes_read += read_iteration; ++buffer_idx; - buffer_idx %= num_buffers; + buffer_idx %= n_buffers; } } else @@ -4182,7 +4364,7 @@ struct llama_model_loader { #if defined(GGML_USE_CUDA) // free temporary resources used for async cuda uploads if (cuda_backend) { - for (size_t idx = 0; idx < num_buffers;++idx) { + for (size_t idx = 0; idx < n_buffers;++idx) { ggml_backend_event_synchronize(events[idx]); ggml_backend_event_free(events[idx]); ggml_backend_buffer_free(host_buffers[idx]); @@ -4304,12 +4486,20 @@ static const char * llama_model_type_name(e_model type) { case MODEL_17M: return "17M"; case MODEL_22M: return "22M"; case MODEL_33M: return "33M"; + case MODEL_60M: return "60M"; case MODEL_70M: return "70M"; + case MODEL_80M: return "80M"; case MODEL_109M: return "109M"; case MODEL_137M: return "137M"; case MODEL_160M: return "160M"; + case MODEL_220M: return "220M"; + case MODEL_250M: return "250M"; + case MODEL_270M: return "270M"; case MODEL_335M: return "335M"; case MODEL_410M: return "410M"; + case MODEL_450M: return "450M"; + case MODEL_770M: return "770M"; + case MODEL_780M: return "780M"; case MODEL_0_5B: return "0.5B"; case MODEL_1B: return "1B"; case MODEL_1_3B: return "1.3B"; @@ -4323,6 +4513,7 @@ static const char * llama_model_type_name(e_model type) { case MODEL_7B: return "7B"; case MODEL_8B: return "8B"; case MODEL_9B: return "9B"; + case MODEL_11B: return "11B"; case MODEL_12B: return "12B"; case MODEL_13B: return "13B"; case MODEL_14B: return "14B"; @@ -4391,20 +4582,18 @@ static void llm_load_hparams( ml.get_key(LLM_KV_GENERAL_NAME, model.name, false); // get hparams kv - ml.get_key(LLM_KV_VOCAB_SIZE, hparams.n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, hparams.n_vocab); + ml.get_key(LLM_KV_VOCAB_SIZE, hparams.n_vocab, false) || ml.get_arr_n(LLM_KV_TOKENIZER_LIST, hparams.n_vocab); // everything past this point is not vocab-related if (hparams.vocab_only) { return; } - ml.get_key(LLM_KV_CONTEXT_LENGTH, hparams.n_ctx_train); - ml.get_key(LLM_KV_EMBEDDING_LENGTH, hparams.n_embd); - ml.get_key(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff); - ml.get_key(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head); - ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer); - ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false); - ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false); + ml.get_key(LLM_KV_CONTEXT_LENGTH, hparams.n_ctx_train); + ml.get_key(LLM_KV_EMBEDDING_LENGTH, hparams.n_embd); + ml.get_key(LLM_KV_BLOCK_COUNT, hparams.n_layer); + ml.get_key(LLM_KV_EXPERT_COUNT, hparams.n_expert, false); + ml.get_key(LLM_KV_EXPERT_USED_COUNT, hparams.n_expert_used, false); GGML_ASSERT(hparams.n_expert <= LLAMA_MAX_EXPERTS); GGML_ASSERT(hparams.n_expert_used <= hparams.n_expert); @@ -4414,9 +4603,18 @@ static void llm_load_hparams( GGML_ASSERT(hparams.n_expert_used == 0); } + // zero-out the per-layer hparams + std::fill(hparams.n_head_arr.begin(), hparams.n_head_arr.end(), 0); + std::fill(hparams.n_head_kv_arr.begin(), hparams.n_head_kv_arr.end(), 0); + std::fill(hparams.n_ff_arr.begin(), hparams.n_ff_arr.end(), 0); + + ml.get_key_or_arr(LLM_KV_FEED_FORWARD_LENGTH, hparams.n_ff_arr, hparams.n_layer); + ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT, hparams.n_head_arr, hparams.n_layer); + // n_head_kv is optional, default to n_head - hparams.n_head_kv = hparams.n_head; - ml.get_key(LLM_KV_ATTENTION_HEAD_COUNT_KV, hparams.n_head_kv, false); + hparams.n_head_kv_arr = hparams.n_head_arr; + + ml.get_key_or_arr(LLM_KV_ATTENTION_HEAD_COUNT_KV, hparams.n_head_kv_arr, hparams.n_layer, false); bool rope_finetuned = false; ml.get_key(LLM_KV_ROPE_SCALING_FINETUNED, rope_finetuned, false); @@ -4444,27 +4642,32 @@ static void llm_load_hparams( ml.get_key(LLM_KV_ROPE_SCALING_ATTN_FACTOR, hparams.rope_attn_factor, false); - // sanity check for n_rot (optional) - { - hparams.n_rot = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head; + // non-transformer models do not have attention heads + if (hparams.n_head() > 0) { + // sanity check for n_rot (optional) + hparams.n_rot = hparams.n_embd / hparams.n_head(); ml.get_key(LLM_KV_ROPE_DIMENSION_COUNT, hparams.n_rot, false); if (model.arch == LLM_ARCH_LLAMA || model.arch == LLM_ARCH_FALCON) { - if (hparams.n_rot != hparams.n_embd / hparams.n_head) { - throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd / hparams.n_head)); + if (hparams.n_rot != hparams.n_embd / hparams.n_head()) { + throw std::runtime_error(format("invalid n_rot: %u, expected %u", hparams.n_rot, hparams.n_embd / hparams.n_head())); } } // gpt-neox n_rot = rotary_pct * (n_embd / n_head) // gpt-j n_rot = rotary_dim + + hparams.n_embd_head_k = hparams.n_embd / hparams.n_head(); + ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH, hparams.n_embd_head_k, false); + + hparams.n_embd_head_v = hparams.n_embd / hparams.n_head(); + ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH, hparams.n_embd_head_v, false); + } else { + hparams.n_rot = 0; + hparams.n_embd_head_k = 0; + hparams.n_embd_head_v = 0; } - hparams.n_embd_head_k = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head; - ml.get_key(LLM_KV_ATTENTION_KEY_LENGTH, hparams.n_embd_head_k, false); - - hparams.n_embd_head_v = (hparams.n_head == 0) ? 0 : hparams.n_embd / hparams.n_head; - ml.get_key(LLM_KV_ATTENTION_VALUE_LENGTH, hparams.n_embd_head_v, false); - // arch-specific KVs switch (model.arch) { case LLM_ARCH_LLAMA: @@ -4487,7 +4690,7 @@ static void llm_load_hparams( case 40: model.type = e_model::MODEL_13B; break; case 48: model.type = e_model::MODEL_34B; break; case 60: model.type = e_model::MODEL_30B; break; - case 80: model.type = hparams.n_head == hparams.n_head_kv ? e_model::MODEL_65B : e_model::MODEL_70B; break; + case 80: model.type = hparams.n_head() == hparams.n_head_kv() ? e_model::MODEL_65B : e_model::MODEL_70B; break; default: model.type = e_model::MODEL_UNKNOWN; } } @@ -4656,7 +4859,7 @@ static void llm_load_hparams( switch (hparams.n_layer) { case 24: model.type = hparams.n_embd == 1024 ? e_model::MODEL_0_5B : e_model::MODEL_1B; break; case 32: model.type = e_model::MODEL_7B; break; - case 40: model.type = hparams.n_head == 20 ? e_model::MODEL_4B : e_model::MODEL_13B; break; + case 40: model.type = hparams.n_head() == 20 ? e_model::MODEL_4B : e_model::MODEL_13B; break; case 80: model.type = e_model::MODEL_70B; break; default: model.type = e_model::MODEL_UNKNOWN; } @@ -4848,46 +5051,58 @@ static void llm_load_hparams( default: model.type = e_model::MODEL_UNKNOWN; } } break; + case LLM_ARCH_OPENELM: + { + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + + switch (hparams.n_layer) { + case 16: model.type = e_model::MODEL_270M; break; + case 20: model.type = e_model::MODEL_450M; break; + case 28: model.type = e_model::MODEL_1B; break; + case 36: model.type = e_model::MODEL_3B; break; + default: model.type = e_model::MODEL_UNKNOWN; + } + } break; case LLM_ARCH_GPTNEOX: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); ml.get_key(LLM_KV_USE_PARALLEL_RESIDUAL, hparams.use_par_res); switch (hparams.n_layer) { case 6: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 512: model.type = e_model::MODEL_14M; break; case 2048: model.type = e_model::MODEL_70M; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 12: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 3072: model.type = e_model::MODEL_160M; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 16: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 8192: model.type = e_model::MODEL_1B; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 24: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 4096: model.type = e_model::MODEL_410M; break; case 8192: model.type = e_model::MODEL_1_4B; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 32: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 10240: model.type = e_model::MODEL_2_8B; break; case 16384: model.type = e_model::MODEL_6_9B; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 36: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 20480: model.type = e_model::MODEL_12B; break; default: model.type = e_model::MODEL_UNKNOWN; } break; case 44: - switch (hparams.n_ff) { + switch (hparams.n_ff()) { case 24576: model.type = e_model::MODEL_20B; break; default: model.type = e_model::MODEL_UNKNOWN; } break; @@ -4945,6 +5160,38 @@ static void llm_load_hparams( default: model.type = e_model::MODEL_UNKNOWN; } } break; + case LLM_ARCH_T5: + { + ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps); + ml.get_key(LLM_KV_ATTENTION_RELATIVE_BUCKETS_COUNT, hparams.n_rel_attn_bkts); + + uint32_t dec_start_token_id; + if (ml.get_key(LLM_KV_DECODER_START_TOKEN_ID, dec_start_token_id, false)) { + hparams.dec_start_token_id = dec_start_token_id; + } + + switch (hparams.n_layer) { + case 6: model.type = e_model::MODEL_60M; break; // t5-small + case 8: model.type = e_model::MODEL_80M; break; // flan-t5-small + case 12: + switch (hparams.n_ff()) { + case 3072: model.type = e_model::MODEL_220M; break; // t5-base + case 2048: model.type = e_model::MODEL_250M; break; // flan-t5-base + default: model.type = e_model::MODEL_UNKNOWN; + } break; + case 24: + switch (hparams.n_ff()) { + case 4096: model.type = e_model::MODEL_770M; break; // t5-large + case 2816: model.type = e_model::MODEL_780M; break; // flan-t5-large + case 16384: model.type = e_model::MODEL_3B; break; // t5-3b + case 5120: model.type = e_model::MODEL_3B; break; // flan-t5-xl + case 65536: model.type = e_model::MODEL_11B; break; // t5-11b + case 10240: model.type = e_model::MODEL_11B; break; // flan-t5-xxl + default: model.type = e_model::MODEL_UNKNOWN; + } break; + default: model.type = e_model::MODEL_UNKNOWN; + } + } break; case LLM_ARCH_JAIS: { ml.get_key(LLM_KV_ATTENTION_LAYERNORM_EPS, hparams.f_norm_eps); @@ -5017,11 +5264,6 @@ static void llm_load_vocab( vocab.special_pad_id = -1; vocab.special_cls_id = -1; vocab.special_mask_id = -1; - - const int add_space_prefix_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_ADD_PREFIX).c_str()); - if (add_space_prefix_keyidx != -1) { - vocab.tokenizer_add_space_prefix = gguf_get_val_bool(ctx, add_space_prefix_keyidx); - } // The default value of add_space_prefix is true. } else if (tokenizer_model == "bert") { vocab.type = LLAMA_VOCAB_TYPE_WPM; @@ -5033,15 +5275,9 @@ static void llm_load_vocab( vocab.special_pad_id = 0; vocab.special_cls_id = 101; vocab.special_mask_id = 103; - vocab.tokenizer_add_space_prefix = false; } else if (tokenizer_model == "gpt2") { vocab.type = LLAMA_VOCAB_TYPE_BPE; - const int add_space_prefix_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_ADD_PREFIX).c_str()); - if (add_space_prefix_keyidx != -1) { - vocab.tokenizer_add_space_prefix = gguf_get_val_bool(ctx, add_space_prefix_keyidx); - } - // read bpe merges and populate bpe ranks const int merges_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_MERGES).c_str()); if (merges_keyidx == -1) { @@ -5118,6 +5354,8 @@ static void llm_load_vocab( // for now, only BPE models have pre-tokenizers if (vocab.type == LLAMA_VOCAB_TYPE_BPE) { + vocab.tokenizer_add_space_prefix = false; + vocab.tokenizer_clean_spaces = true; if (tokenizer_pre.empty()) { LLAMA_LOG_WARN("%s: missing pre-tokenizer type, using: 'default'\n", __func__); LLAMA_LOG_WARN("%s: \n", __func__); @@ -5139,9 +5377,11 @@ static void llm_load_vocab( } else if ( tokenizer_pre == "deepseek-llm") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_LLM; + vocab.tokenizer_clean_spaces = false; } else if ( tokenizer_pre == "deepseek-coder") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEEPSEEK_CODER; + vocab.tokenizer_clean_spaces = false; } else if ( tokenizer_pre == "falcon") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_FALCON; @@ -5153,6 +5393,7 @@ static void llm_load_vocab( vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STARCODER; } else if ( tokenizer_pre == "gpt-2" || + tokenizer_pre == "phi-2" || tokenizer_pre == "jina-es" || tokenizer_pre == "jina-de" || tokenizer_pre == "jina-v2-es" || @@ -5168,6 +5409,7 @@ static void llm_load_vocab( } else if ( tokenizer_pre == "qwen2") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_QWEN2; + vocab.tokenizer_clean_spaces = false; } else if ( tokenizer_pre == "stablelm2") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_STABLELM2; @@ -5183,6 +5425,7 @@ static void llm_load_vocab( } else if ( tokenizer_pre == "poro-chat") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_PORO; + vocab.tokenizer_clean_spaces = false; } else if ( tokenizer_pre == "chatglm-bpe") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_CHATGLM4; @@ -5190,6 +5433,7 @@ static void llm_load_vocab( } else if ( tokenizer_pre == "viking") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_VIKING; + vocab.tokenizer_clean_spaces = false; } else if ( tokenizer_pre == "jais") { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_JAIS; @@ -5198,10 +5442,14 @@ static void llm_load_vocab( } } else if (vocab.type == LLAMA_VOCAB_TYPE_SPM) { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; + vocab.tokenizer_add_space_prefix = true; + vocab.tokenizer_clean_spaces = false; vocab.tokenizer_add_bos = true; vocab.tokenizer_add_eos = false; } else if (vocab.type == LLAMA_VOCAB_TYPE_WPM) { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; + vocab.tokenizer_add_space_prefix = false; + vocab.tokenizer_clean_spaces = true; vocab.tokenizer_add_bos = true; vocab.tokenizer_add_eos = false; } else if (vocab.type == LLAMA_VOCAB_TYPE_UGM) { @@ -5211,6 +5459,11 @@ static void llm_load_vocab( } else { vocab.type_pre = LLAMA_VOCAB_PRE_TYPE_DEFAULT; } + + const int add_space_prefix_keyidx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_ADD_PREFIX).c_str()); + if (add_space_prefix_keyidx != -1) { + vocab.tokenizer_add_space_prefix = gguf_get_val_bool(ctx, add_space_prefix_keyidx); + } } const int token_idx = gguf_find_key(ctx, kv(LLM_KV_TOKENIZER_LIST).c_str()); @@ -5391,7 +5644,7 @@ static void llm_load_vocab( } } - std::sort( vocab.cache_special_tokens.begin(), vocab.cache_special_tokens.end(), + std::sort(vocab.cache_special_tokens.begin(), vocab.cache_special_tokens.end(), [&] (const llama_vocab::id a, const llama_vocab::id b) { return vocab.id_to_token[a].text.size() > vocab.id_to_token[b].text.size(); } @@ -5477,44 +5730,78 @@ static void llm_load_print_meta(llama_model_loader & ml, llama_model & model) { const char * rope_scaling_type = LLAMA_ROPE_SCALING_TYPES.at(hparams.rope_scaling_type_train); + auto print_f = [](const std::function & f, uint32_t n) { + bool is_var = false; + + std::vector v; + for (uint32_t i = 0; i < n; ++i) { + v.push_back(f(i)); + if (v[i] != v[0]) { + is_var = true; + } + } + + std::stringstream ss; + + if (is_var) { + ss << "["; + for (uint32_t i = 0; i < n; ++i) { + ss << v[i]; + if (i < n - 1) { + ss << ", "; + } + } + ss << "]"; + } else { + ss << v[0]; + } + + return ss.str(); + }; + // hparams LLAMA_LOG_INFO("%s: format = %s\n", __func__, llama_file_version_name(ml.fver)); LLAMA_LOG_INFO("%s: arch = %s\n", __func__, LLM_ARCH_NAMES.at(model.arch)); LLAMA_LOG_INFO("%s: vocab type = %s\n", __func__, llama_model_vocab_type_name(vocab.type)); LLAMA_LOG_INFO("%s: n_vocab = %u\n", __func__, hparams.n_vocab); LLAMA_LOG_INFO("%s: n_merges = %u\n", __func__, (int) vocab.bpe_ranks.size()); - LLAMA_LOG_INFO("%s: n_ctx_train = %u\n", __func__, hparams.n_ctx_train); - LLAMA_LOG_INFO("%s: n_embd = %u\n", __func__, hparams.n_embd); - LLAMA_LOG_INFO("%s: n_head = %u\n", __func__, hparams.n_head); - LLAMA_LOG_INFO("%s: n_head_kv = %u\n", __func__, hparams.n_head_kv); - LLAMA_LOG_INFO("%s: n_layer = %u\n", __func__, hparams.n_layer); - LLAMA_LOG_INFO("%s: n_rot = %u\n", __func__, hparams.n_rot); - LLAMA_LOG_INFO("%s: n_swa = %u\n", __func__, hparams.n_swa); - LLAMA_LOG_INFO("%s: n_embd_head_k = %u\n", __func__, hparams.n_embd_head_k); - LLAMA_LOG_INFO("%s: n_embd_head_v = %u\n", __func__, hparams.n_embd_head_v); - LLAMA_LOG_INFO("%s: n_gqa = %u\n", __func__, hparams.n_gqa()); - LLAMA_LOG_INFO("%s: n_embd_k_gqa = %u\n", __func__, hparams.n_embd_k_gqa()); - LLAMA_LOG_INFO("%s: n_embd_v_gqa = %u\n", __func__, hparams.n_embd_v_gqa()); - LLAMA_LOG_INFO("%s: f_norm_eps = %.1e\n", __func__, hparams.f_norm_eps); - LLAMA_LOG_INFO("%s: f_norm_rms_eps = %.1e\n", __func__, hparams.f_norm_rms_eps); - LLAMA_LOG_INFO("%s: f_clamp_kqv = %.1e\n", __func__, hparams.f_clamp_kqv); - LLAMA_LOG_INFO("%s: f_max_alibi_bias = %.1e\n", __func__, hparams.f_max_alibi_bias); - LLAMA_LOG_INFO("%s: f_logit_scale = %.1e\n", __func__, hparams.f_logit_scale); - LLAMA_LOG_INFO("%s: n_ff = %u\n", __func__, hparams.n_ff); - LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert); - LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used); - LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn); - LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type); - LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type); - LLAMA_LOG_INFO("%s: rope scaling = %s\n", __func__, rope_scaling_type); - LLAMA_LOG_INFO("%s: freq_base_train = %.1f\n", __func__, hparams.rope_freq_base_train); - LLAMA_LOG_INFO("%s: freq_scale_train = %g\n", __func__, hparams.rope_freq_scale_train); - LLAMA_LOG_INFO("%s: n_ctx_orig_yarn = %u\n", __func__, hparams.n_ctx_orig_yarn); - LLAMA_LOG_INFO("%s: rope_finetuned = %s\n", __func__, hparams.rope_finetuned ? "yes" : "unknown"); - LLAMA_LOG_INFO("%s: ssm_d_conv = %u\n", __func__, hparams.ssm_d_conv); - LLAMA_LOG_INFO("%s: ssm_d_inner = %u\n", __func__, hparams.ssm_d_inner); - LLAMA_LOG_INFO("%s: ssm_d_state = %u\n", __func__, hparams.ssm_d_state); - LLAMA_LOG_INFO("%s: ssm_dt_rank = %u\n", __func__, hparams.ssm_dt_rank); + LLAMA_LOG_INFO("%s: vocab_only = %d\n", __func__, hparams.vocab_only); + + if (!hparams.vocab_only) { + LLAMA_LOG_INFO("%s: n_ctx_train = %u\n", __func__, hparams.n_ctx_train); + LLAMA_LOG_INFO("%s: n_embd = %u\n", __func__, hparams.n_embd); + LLAMA_LOG_INFO("%s: n_layer = %u\n", __func__, hparams.n_layer); + LLAMA_LOG_INFO("%s: n_head = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_head_kv = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_head_kv(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_rot = %u\n", __func__, hparams.n_rot); + LLAMA_LOG_INFO("%s: n_swa = %u\n", __func__, hparams.n_swa); + LLAMA_LOG_INFO("%s: n_embd_head_k = %u\n", __func__, hparams.n_embd_head_k); + LLAMA_LOG_INFO("%s: n_embd_head_v = %u\n", __func__, hparams.n_embd_head_v); + LLAMA_LOG_INFO("%s: n_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_gqa(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_embd_k_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_k_gqa(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_embd_v_gqa = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_embd_v_gqa(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: f_norm_eps = %.1e\n", __func__, hparams.f_norm_eps); + LLAMA_LOG_INFO("%s: f_norm_rms_eps = %.1e\n", __func__, hparams.f_norm_rms_eps); + LLAMA_LOG_INFO("%s: f_clamp_kqv = %.1e\n", __func__, hparams.f_clamp_kqv); + LLAMA_LOG_INFO("%s: f_max_alibi_bias = %.1e\n", __func__, hparams.f_max_alibi_bias); + LLAMA_LOG_INFO("%s: f_logit_scale = %.1e\n", __func__, hparams.f_logit_scale); + LLAMA_LOG_INFO("%s: n_ff = %s\n", __func__, print_f([&](uint32_t il) { return hparams.n_ff(il); }, hparams.n_layer).c_str()); + LLAMA_LOG_INFO("%s: n_expert = %u\n", __func__, hparams.n_expert); + LLAMA_LOG_INFO("%s: n_expert_used = %u\n", __func__, hparams.n_expert_used); + LLAMA_LOG_INFO("%s: causal attn = %d\n", __func__, hparams.causal_attn); + LLAMA_LOG_INFO("%s: pooling type = %d\n", __func__, hparams.pooling_type); + LLAMA_LOG_INFO("%s: rope type = %d\n", __func__, hparams.rope_type); + LLAMA_LOG_INFO("%s: rope scaling = %s\n", __func__, rope_scaling_type); + LLAMA_LOG_INFO("%s: freq_base_train = %.1f\n", __func__, hparams.rope_freq_base_train); + LLAMA_LOG_INFO("%s: freq_scale_train = %g\n", __func__, hparams.rope_freq_scale_train); + LLAMA_LOG_INFO("%s: n_ctx_orig_yarn = %u\n", __func__, hparams.n_ctx_orig_yarn); + LLAMA_LOG_INFO("%s: rope_finetuned = %s\n", __func__, hparams.rope_finetuned ? "yes" : "unknown"); + LLAMA_LOG_INFO("%s: ssm_d_conv = %u\n", __func__, hparams.ssm_d_conv); + LLAMA_LOG_INFO("%s: ssm_d_inner = %u\n", __func__, hparams.ssm_d_inner); + LLAMA_LOG_INFO("%s: ssm_d_state = %u\n", __func__, hparams.ssm_d_state); + LLAMA_LOG_INFO("%s: ssm_dt_rank = %u\n", __func__, hparams.ssm_dt_rank); + } + LLAMA_LOG_INFO("%s: model type = %s\n", __func__, llama_model_type_name(model.type)); LLAMA_LOG_INFO("%s: model ftype = %s\n", __func__, llama_model_ftype_name(model.ftype).c_str()); if (ml.n_elements >= 1e12) { @@ -5594,8 +5881,8 @@ static bool llm_load_tensors( model.main_gpu = main_gpu; model.n_gpu_layers = n_gpu_layers; - const int64_t n_layer = hparams.n_layer; - const int64_t i_gpu_start = std::max((int64_t) hparams.n_layer - n_gpu_layers, (int64_t) 0); + const int n_layer = hparams.n_layer; + const int i_gpu_start = std::max((int) hparams.n_layer - n_gpu_layers, (int) 0); bool use_mmap_buffer = true; // there is very little benefit to offloading the input layer, so always keep it on the CPU @@ -5605,7 +5892,7 @@ static bool llm_load_tensors( model.buft_layer.resize(n_layer); // assign cpu layers - for (int64_t i = 0; i < i_gpu_start; ++i) { + for (int i = 0; i < i_gpu_start; ++i) { model.buft_layer[i] = llama_default_buffer_type_cpu(true); } @@ -5635,7 +5922,7 @@ static bool llm_load_tensors( // assign the repeating layers to the devices according to the splits int act_gpu_layers = std::min(n_gpu_layers, (int)n_layer + 1); - for (int64_t i = i_gpu_start; i < n_layer; ++i) { + for (int i = i_gpu_start; i < n_layer; ++i) { int layer_gpu = std::upper_bound(splits.begin(), splits.begin() + device_count, float(i - i_gpu_start)/act_gpu_layers) - splits.begin(); model.buft_layer[i] = llama_default_buffer_type_offload(model, layer_gpu); } @@ -5655,7 +5942,7 @@ static bool llm_load_tensors( split_buft = llama_default_buffer_type_offload(model, main_gpu); } // assign the repeating layers - for (int64_t i = i_gpu_start; i < n_layer; ++i) { + for (int i = i_gpu_start; i < n_layer; ++i) { model.buft_layer[i] = { split_buft, llama_default_buffer_type_offload(model, main_gpu) @@ -5678,7 +5965,7 @@ static bool llm_load_tensors( buft_layer_count[model.buft_input.buft_matrix]++; buft_layer_count[model.buft_output.buft]++; buft_layer_count[model.buft_output.buft_matrix]++; - for (int64_t i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer; ++i) { buft_layer_count[model.buft_layer[i].buft]++; buft_layer_count[model.buft_layer[i].buft_matrix]++; } @@ -5708,15 +5995,21 @@ static bool llm_load_tensors( // create tensors for the weights { - const int64_t n_embd = hparams.n_embd; - const int64_t n_embd_head = (hparams.n_head == 0) ? 0 : n_embd / hparams.n_head; - const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); - const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); - const int64_t n_embd_gqa = n_embd_v_gqa; - const int64_t n_vocab = hparams.n_vocab; - const int64_t n_vocab_type = hparams.n_vocab_type; - const int64_t n_ff = hparams.n_ff; - const int64_t n_expert = hparams.n_expert; + // note: cast to int64_t since we will use these for the tensor dimensions + const int64_t n_head = hparams.n_head(); + const int64_t n_head_kv = hparams.n_head_kv(); + const int64_t n_embd = hparams.n_embd; + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); + const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); + const int64_t n_embd_head_k = hparams.n_embd_head_k; + const int64_t n_embd_head_v = hparams.n_embd_head_v; + const int64_t n_ff = hparams.n_ff(); + const int64_t n_embd_gqa = n_embd_v_gqa; + const int64_t n_vocab = hparams.n_vocab; + const int64_t n_vocab_type = hparams.n_vocab_type; + const int64_t n_expert = hparams.n_expert; + const int64_t n_expert_used = hparams.n_expert_used; + const int64_t n_ctx_train = hparams.n_ctx_train; if (n_expert > 0 && hparams.n_expert_used == 0) { throw std::runtime_error("model has expert layers but no expert layers are used"); @@ -5725,8 +6018,9 @@ static bool llm_load_tensors( ggml_context * ctx_input = ctx_map.at(model.buft_input.buft); ggml_context * ctx_output = ctx_map.at(model.buft_output.buft); ggml_context * ctx_output_split = ctx_map.at(model.buft_output.buft_matrix); - auto ctx_for_layer = [&](int i) { return ctx_map.at(model.buft_layer[i].buft); }; - auto ctx_for_layer_split = [&](int i) { return ctx_map.at(model.buft_layer[i].buft_matrix); }; + + auto ctx_for_layer = [&](int i) { return ctx_map.at(model.buft_layer[i].buft); }; + auto ctx_for_layer_split = [&](int i) { return ctx_map.at(model.buft_layer[i].buft_matrix); }; model.layers.resize(n_layer); @@ -5741,7 +6035,8 @@ static bool llm_load_tensors( // output { model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); - model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + // if output is NULL, init from the input tok embed if (model.output == NULL) { model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); @@ -5824,6 +6119,7 @@ static bool llm_load_tensors( { model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + // if output is NULL, init from the input tok embed if (model.output == NULL) { model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); @@ -5847,9 +6143,9 @@ static bool llm_load_tensors( layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); - layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}); - + layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}); layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}, llama_model_loader::TENSOR_NOT_REQUIRED); + if (layer.ffn_gate_exps) { layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), { n_ff, n_embd, n_expert}); layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), {n_embd, n_ff, n_expert}); @@ -5901,12 +6197,12 @@ static bool llm_load_tensors( auto & layer = model.layers[i]; - layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); + layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}); layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}); + layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}); layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}); layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff, n_expert}); @@ -5948,10 +6244,10 @@ static bool llm_load_tensors( // output { - model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); - model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}); + model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); + model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}); - model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); if (!model.output) { model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); // needs to be on GPU } @@ -5979,7 +6275,7 @@ static bool llm_load_tensors( case LLM_ARCH_STARCODER: { model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); - model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train}); + model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, n_ctx_train}); // output { @@ -6005,8 +6301,8 @@ static bool llm_load_tensors( layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}); layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); + layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}); @@ -6014,8 +6310,8 @@ static bool llm_load_tensors( layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); - layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); - layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); } } break; case LLM_ARCH_BERT: @@ -6023,8 +6319,9 @@ static bool llm_load_tensors( { model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); model.type_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_vocab_type}); + if (model.arch == LLM_ARCH_BERT) { - model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train}); + model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, n_ctx_train}); } model.tok_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}); @@ -6037,31 +6334,30 @@ static bool llm_load_tensors( auto & layer = model.layers[i]; if (model.arch == LLM_ARCH_BERT) { - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); - layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); + layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}); - layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); - layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}); + layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); + layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}); - layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); - layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}); + layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); + layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}); } else { layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}); } - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}); layer.attn_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i), {n_embd}); - layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); - layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); if (model.arch == LLM_ARCH_BERT) { - layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); - layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); - - layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); + layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); + layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); + layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); } else { layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); } @@ -6072,8 +6368,9 @@ static bool llm_load_tensors( } break; case LLM_ARCH_JINA_BERT_V2: { - model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // word_embeddings - model.type_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_vocab_type}); //token_type_embeddings + model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); // word_embeddings + model.type_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_TYPES, "weight"), {n_embd, n_vocab_type}); // token_type_embeddings + model.tok_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "weight"), {n_embd}); // LayerNorm model.tok_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD_NORM, "bias"), {n_embd}); //LayerNorm bias @@ -6083,38 +6380,38 @@ static bool llm_load_tensors( auto & layer = model.layers[i]; // JinaBertLayer - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); - layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); + layer.bq = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "bias", i), {n_embd}); layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.attn_q_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.attn_q_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); - layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}); + layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); + layer.bk = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "bias", i), {n_embd_gqa}); layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.attn_k_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.attn_k_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); - layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}); + layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); + layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); //output_dens - layer.bo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); //output_dens + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); //output_dens + layer.bo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); //output_dens layer.attn_out_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "weight", i), {n_embd}); //output_norm - layer.attn_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i), {n_embd}); + layer.attn_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT_NORM, "bias", i), {n_embd}); layer.attn_norm_2 = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM_2, "weight", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.attn_norm_2_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM_2, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.attn_norm_2_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM_2, "bias", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); - layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); - layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); - layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); + layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); + layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); - layer.layer_out_norm = ml.create_tensor(ctx_split, tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}); - layer.layer_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i), {n_embd}); + layer.layer_out_norm = ml.create_tensor(ctx_split, tn(LLM_TENSOR_LAYER_OUT_NORM, "weight", i), {n_embd}); + layer.layer_out_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_LAYER_OUT_NORM, "bias", i), {n_embd}); } } break; case LLM_ARCH_BLOOM: @@ -6137,35 +6434,35 @@ static bool llm_load_tensors( auto & layer = model.layers[i]; layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); - layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}); + layer.attn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "bias", i), {n_embd}); layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}); - layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}); + layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}); layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); + layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); - layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}); + layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}); layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); - layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); + layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); - layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); - layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); } } break; case LLM_ARCH_MPT: { model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); - model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, n_ctx_train}, llama_model_loader::TENSOR_NOT_REQUIRED); // output { - model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); - model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); + model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); - model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); if (!model.output) { model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); // needs to be on GPU } @@ -6236,8 +6533,8 @@ static bool llm_load_tensors( layer.bv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "bias", i), {n_embd_gqa}, llama_model_loader::TENSOR_NOT_REQUIRED); // optional q and k layernorms, present in StableLM 2 12B - layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head}, llama_model_loader::TENSOR_NOT_REQUIRED); - layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head_kv}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}, llama_model_loader::TENSOR_NOT_REQUIRED); // optional FFN norm, not present in StableLM 2 12B which uses parallel residual layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}, llama_model_loader::TENSOR_NOT_REQUIRED); @@ -6348,21 +6645,23 @@ static bool llm_load_tensors( layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}); - GGML_ASSERT(hparams.n_expert > 0); - GGML_ASSERT(hparams.n_expert_used > 0); + GGML_ASSERT(n_expert > 0); + GGML_ASSERT(n_expert_used > 0); // MoE branch - auto n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / hparams.n_expert_used; + const int64_t n_ff_exp = hparams.n_ff_exp ? hparams.n_ff_exp : n_ff / n_expert_used; + layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}); layer.ffn_down_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}); layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}); // Shared expert branch - auto n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff; + const int64_t n_ff_shexp = hparams.n_ff_shexp ? hparams.n_ff_shexp : n_ff; + layer.ffn_gate_inp_shexp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP_SHEXP, "weight", i), {n_embd}); - layer.ffn_gate_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}); - layer.ffn_down_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}); - layer.ffn_up_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_shexp}); + layer.ffn_gate_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), { n_embd, n_ff_shexp}); + layer.ffn_down_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}); + layer.ffn_up_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), { n_embd, n_ff_shexp}); } } break; case LLM_ARCH_PHI2: @@ -6412,6 +6711,8 @@ static bool llm_load_tensors( } break; case LLM_ARCH_PHI3: { + const int64_t n_embd_head = n_embd / n_head; + model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), { n_embd, n_vocab }); // output @@ -6421,8 +6722,8 @@ static bool llm_load_tensors( } for (int i = 0; i < n_layer; ++i) { - ggml_context* ctx_layer = ctx_for_layer(i); - ggml_context* ctx_split = ctx_for_layer_split(i); + ggml_context * ctx_layer = ctx_for_layer(i); + ggml_context * ctx_split = ctx_for_layer_split(i); auto & layer = model.layers[i]; @@ -6471,7 +6772,7 @@ static bool llm_load_tensors( case LLM_ARCH_GPT2: { model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); - model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, hparams.n_ctx_train}); + model.pos_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_POS_EMBD, "weight"), {n_embd, n_ctx_train}); // output { @@ -6608,12 +6909,7 @@ static bool llm_load_tensors( model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading - const int64_t n_ff = hparams.n_ff; - const int64_t n_embd_head_k = hparams.n_embd_head_k; - const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); - const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); - - for (uint32_t i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer; ++i) { ggml_context * ctx_layer = ctx_for_layer(i); ggml_context * ctx_split = ctx_for_layer_split(i); @@ -6621,10 +6917,10 @@ static bool llm_load_tensors( layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * hparams.n_head}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * n_head}); layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}); layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * hparams.n_head, n_embd}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); @@ -6640,12 +6936,7 @@ static bool llm_load_tensors( model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); // same as tok_embd, duplicated to allow offloading - const int64_t n_ff = hparams.n_ff; - const int64_t n_embd_head_k = hparams.n_embd_head_k; - const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); - const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); - - for (uint32_t i = 0; i < n_layer; ++i) { + for (int i = 0; i < n_layer; ++i) { ggml_context * ctx_layer = ctx_for_layer(i); ggml_context * ctx_split = ctx_for_layer_split(i); @@ -6653,10 +6944,10 @@ static bool llm_load_tensors( layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * hparams.n_head}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_head_k * n_head}); layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}); layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * hparams.n_head, n_embd}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}); layer.attn_post_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); @@ -6720,6 +7011,7 @@ static bool llm_load_tensors( const int64_t d_inner = hparams.ssm_d_inner; const int64_t d_state = hparams.ssm_d_state; const int64_t dt_rank = hparams.ssm_dt_rank; + // only an expansion factor of 2 is supported for now GGML_ASSERT(2 * n_embd == d_inner); @@ -6770,15 +7062,20 @@ static bool llm_load_tensors( model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}); } + for (int i = 0; i < n_layer; ++i) { ggml_context * ctx_layer = ctx_for_layer(i); ggml_context * ctx_split = ctx_for_layer_split(i); + auto & layer = model.layers[i]; + layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); + layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}); @@ -6805,8 +7102,8 @@ static bool llm_load_tensors( layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); if (n_layer >= 64){ - layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head}); - layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {hparams.n_embd_head_k, hparams.n_head_kv}); + layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k, n_head}); + layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k, n_head_kv}); } layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); @@ -6842,15 +7139,49 @@ static bool llm_load_tensors( layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}); layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); } } break; + case LLM_ARCH_OPENELM: + { + model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); + + // output + { + model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); + // init output from the input tok embed + model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); + } + + for (int i = 0; i < n_layer; ++i) { + const int64_t n_head = hparams.n_head(i); + const int64_t n_head_qkv = 2*hparams.n_head_kv(i) + n_head; + const int64_t n_ff = hparams.n_ff(i); + + ggml_context * ctx_layer = ctx_for_layer(i); + ggml_context * ctx_split = ctx_for_layer_split(i); + + auto & layer = model.layers[i]; + + layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); + + layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_head_qkv*n_embd_head_k}); + layer.attn_q_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}); + layer.attn_k_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_head*n_embd_head_k, n_embd}); + + layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); + layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); + layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + } + } break; case LLM_ARCH_GPTNEOX: { - model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); + model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); + // output { model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); @@ -6889,8 +7220,9 @@ static bool llm_load_tensors( // output { - model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); - model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}); + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + // if output is NULL, init from the input tok embed if (model.output == NULL) { model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); @@ -6925,13 +7257,16 @@ static bool llm_load_tensors( } break; case LLM_ARCH_DEEPSEEK2: { - bool is_lite = (hparams.n_layer == 27); + const bool is_lite = (hparams.n_layer == 27); - const uint32_t n_embd_head_qk_rope = hparams.n_rot; - const uint32_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot; - const uint32_t q_lora_rank = hparams.n_lora_q; - const uint32_t kv_lora_rank = hparams.n_lora_kv; - const uint32_t n_ff_exp = hparams.n_ff_exp; + const int64_t n_embd_head_qk_rope = hparams.n_rot; + const int64_t n_embd_head_qk_nope = hparams.n_embd_head_k - hparams.n_rot; + + const int64_t q_lora_rank = hparams.n_lora_q; + const int64_t kv_lora_rank = hparams.n_lora_kv; + + const int64_t n_ff_exp = hparams.n_ff_exp; + const int64_t n_expert_shared = hparams.n_expert_shared; model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); @@ -6951,29 +7286,31 @@ static bool llm_load_tensors( if (!is_lite) { layer.attn_q_a_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q_A_NORM, "weight", i), {q_lora_rank}); } + layer.attn_kv_a_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_KV_A_NORM, "weight", i), {kv_lora_rank}); if (!is_lite) { - layer.wq_a = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}); - layer.wq_b = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, hparams.n_head * hparams.n_embd_head_k}); + layer.wq_a = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q_A, "weight", i), {n_embd, q_lora_rank}); + layer.wq_b = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q_B, "weight", i), {q_lora_rank, n_head * n_embd_head_k}); } else { - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}); } - layer.wkv_a_mqa = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + n_embd_head_qk_rope}); - layer.wkv_b = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_KV_B, "weight", i), {kv_lora_rank, hparams.n_head * (n_embd_head_qk_nope + hparams.n_embd_head_v)}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {hparams.n_head * hparams.n_embd_head_v, n_embd}); + + layer.wkv_a_mqa = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_KV_A_MQA, "weight", i), {n_embd, kv_lora_rank + (n_embd_head_qk_rope)}); + layer.wkv_b = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_KV_B, "weight", i), {kv_lora_rank, n_head * (n_embd_head_qk_nope + n_embd_head_v)}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), { n_head * ( n_embd_head_v), n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); - if ((uint32_t) i < hparams.n_layer_dense_lead) { + if (i < (int) hparams.n_layer_dense_lead) { layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), { n_ff, n_embd}); layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); } else { layer.ffn_gate_inp = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}); - GGML_ASSERT(hparams.n_expert > 0); - GGML_ASSERT(hparams.n_expert_used > 0); + GGML_ASSERT(n_expert > 0); + GGML_ASSERT(n_expert_used > 0); // MoE branch layer.ffn_gate_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}); @@ -6981,9 +7318,9 @@ static bool llm_load_tensors( layer.ffn_up_exps = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_EXPS, "weight", i), { n_embd, n_ff_exp, n_expert}); // Shared expert branch - layer.ffn_gate_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * hparams.n_expert_shared}); - layer.ffn_down_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_exp * hparams.n_expert_shared, n_embd}); - layer.ffn_up_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_exp * hparams.n_expert_shared}); + layer.ffn_gate_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}); + layer.ffn_down_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), { n_ff_exp * n_expert_shared, n_embd}); + layer.ffn_up_shexp = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP_SHEXP, "weight", i), {n_embd, n_ff_exp * n_expert_shared}); } } } break; @@ -7002,27 +7339,87 @@ static bool llm_load_tensors( auto & layer = model.layers[i]; - layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); + layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_NORM, "weight", i), {n_embd}); layer.attn_sub_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_SUB_NORM, "weight", i), {n_embd}); - layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); - layer.wq_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "scale", i), {1}); - layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); - layer.wk_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "scale", i), {1}); - layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); - layer.wv_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "scale", i), {1}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.wo_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "scale", i), {1}); + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_Q, "weight", i), {n_embd, n_embd}); + layer.wq_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_Q, "scale", i), {1}); + layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_K, "weight", i), {n_embd, n_embd_gqa}); + layer.wk_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_K, "scale", i), {1}); + layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_V, "weight", i), {n_embd, n_embd_gqa}); + layer.wv_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_V, "scale", i), {1}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); + layer.wo_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "scale", i), {1}); - layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); + layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); layer.ffn_sub_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_SUB_NORM, "weight", i), {n_ff}); - layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); - layer.ffn_gate_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE, "scale", i), {1}); - layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); - layer.ffn_down_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "scale", i), {1}); - layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); - layer.ffn_up_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "scale", i), {1}); + layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); + layer.ffn_gate_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE, "scale", i), {1}); + layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); + layer.ffn_down_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "scale", i), {1}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); + layer.ffn_up_scale = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "scale", i), {1}); + } + } break; + case LLM_ARCH_T5: + { + const auto n_rel_attn_bkts = hparams.n_rel_attn_bkts; + + model.tok_embd = ml.create_tensor(ctx_input, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}); + + // output + { + model.output_norm_enc = ml.create_tensor(ctx_output, tn(LLM_TENSOR_ENC_OUTPUT_NORM, "weight"), {n_embd}); + model.output_norm = ml.create_tensor(ctx_output, tn(LLM_TENSOR_DEC_OUTPUT_NORM, "weight"), {n_embd}); + + model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_NOT_REQUIRED); + // if output is NULL, init from the input tok embed + if (model.output == NULL) { + model.output = ml.create_tensor(ctx_output, tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, llama_model_loader::TENSOR_DUPLICATED); + } + } + + for (int i = 0; i < n_layer; ++i) { + ggml_context * ctx_layer = ctx_for_layer(i); + ggml_context * ctx_split = ctx_for_layer_split(i); + + auto & layer = model.layers[i]; + + layer.attn_norm_enc = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ENC_ATTN_NORM, "weight", i), {n_embd}); + layer.attn_rel_b_enc = ml.create_tensor(ctx_input, tn(LLM_TENSOR_ENC_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, llama_model_loader::TENSOR_NOT_REQUIRED); + + layer.wq_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wk_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wv_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}); + layer.wo_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}); + + layer.ffn_norm_enc = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ENC_FFN_NORM, "weight", i), {n_embd}); + layer.ffn_gate_enc = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ENC_FFN_GATE, "weight", i), {n_embd, n_ff}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.ffn_down_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_FFN_DOWN, "weight", i), { n_ff, n_embd}); + layer.ffn_up_enc = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ENC_FFN_UP, "weight", i), {n_embd, n_ff}); + + layer.attn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_DEC_ATTN_NORM, "weight", i), {n_embd}); + layer.attn_rel_b = ml.create_tensor(ctx_input, tn(LLM_TENSOR_DEC_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, llama_model_loader::TENSOR_NOT_REQUIRED); + + layer.wq = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wk = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}); + + layer.attn_norm_cross = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_DEC_CROSS_ATTN_NORM, "weight", i), {n_embd}); + // this tensor seems to be unused in HF transformers implementation + layer.attn_rel_b_cross = ml.create_tensor(ctx_input, tn(LLM_TENSOR_DEC_CROSS_ATTN_REL_B, "weight", i), {n_head, n_rel_attn_bkts}, llama_model_loader::TENSOR_NOT_REQUIRED); + + layer.wq_cross = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_CROSS_ATTN_Q, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wk_cross = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_CROSS_ATTN_K, "weight", i), {n_embd, n_embd_k_gqa}); + layer.wv_cross = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_CROSS_ATTN_V, "weight", i), {n_embd, n_embd_v_gqa}); + layer.wo_cross = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_CROSS_ATTN_OUT, "weight", i), {n_embd_v_gqa, n_embd}); + + layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_DEC_FFN_NORM, "weight", i), {n_embd}); + layer.ffn_gate = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_DEC_FFN_GATE, "weight", i), {n_embd, n_ff}, llama_model_loader::TENSOR_NOT_REQUIRED); + layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_FFN_DOWN, "weight", i), { n_ff, n_embd}); + layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_DEC_FFN_UP, "weight", i), {n_embd, n_ff}); } } break; case LLM_ARCH_JAIS: @@ -7035,6 +7432,7 @@ static bool llm_load_tensors( model.output_norm_b = ml.create_tensor(ctx_output, tn(LLM_TENSOR_OUTPUT_NORM, "bias"), {n_embd}); model.output = ml.create_tensor(ctx_output_split, tn(LLM_TENSOR_OUTPUT, "weight"), {n_embd, n_vocab}); } + for (int i = 0; i < n_layer; ++i) { ggml_context * ctx_layer = ctx_for_layer(i); ggml_context * ctx_split = ctx_for_layer_split(i); @@ -7047,8 +7445,8 @@ static bool llm_load_tensors( layer.wqkv = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_QKV, "weight", i), {n_embd, n_embd + 2*n_embd_gqa}); layer.bqkv = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_QKV, "bias", i), {n_embd + 2*n_embd_gqa}); - layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); - layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); + layer.wo = ml.create_tensor(ctx_split, tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd, n_embd}); + layer.bo = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_ATTN_OUT, "bias", i), {n_embd}); layer.ffn_norm = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "weight", i), {n_embd}); layer.ffn_norm_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_NORM, "bias", i), {n_embd}); @@ -7056,8 +7454,8 @@ static bool llm_load_tensors( layer.ffn_down = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}); layer.ffn_down_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_DOWN, "bias", i), {n_embd}); - layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); - layer.ffn_gate_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}); + layer.ffn_gate = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}); + layer.ffn_gate_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_GATE, "bias", i), {n_ff}); layer.ffn_up = ml.create_tensor(ctx_split, tn(LLM_TENSOR_FFN_UP, "weight", i), {n_embd, n_ff}); layer.ffn_up_b = ml.create_tensor(ctx_layer, tn(LLM_TENSOR_FFN_UP, "bias", i), {n_ff}); @@ -7372,8 +7770,8 @@ static void llm_build_kv_store( int64_t il) { const int64_t n_ctx = cparams.n_ctx; - const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); - const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(il); + const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(il); GGML_ASSERT(kv.size == n_ctx); @@ -7679,12 +8077,12 @@ static struct ggml_tensor * llm_build_kqv( const llm_build_cb & cb, int il) { const int64_t n_ctx = cparams.n_ctx; - const int64_t n_head = hparams.n_head; - const int64_t n_head_kv = hparams.n_head_kv; + const int64_t n_head = hparams.n_head(il); + const int64_t n_head_kv = hparams.n_head_kv(il); const int64_t n_embd_head_k = hparams.n_embd_head_k; - const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(); + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(il); const int64_t n_embd_head_v = hparams.n_embd_head_v; - const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(); + const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(il); struct ggml_tensor * q = ggml_permute(ctx, q_cur, 0, 2, 1, 3); cb(q, "q", il); @@ -7860,6 +8258,7 @@ struct llm_build_context { const int32_t n_tokens; const int32_t n_kv; // size of KV cache to consider (n_kv <= kv_self.size) const int32_t n_outputs; + const int32_t n_outputs_enc; const int32_t kv_head; // index of where we store new KV data in the cache const int32_t n_ctx_orig; @@ -7890,8 +8289,8 @@ struct llm_build_context { n_layer (hparams.n_layer), n_rot (hparams.n_rot), n_ctx (cparams.n_ctx), - n_head (hparams.n_head), - n_head_kv (hparams.n_head_kv), + n_head (hparams.n_head()), + n_head_kv (hparams.n_head_kv()), n_embd_head_k (hparams.n_embd_head_k), n_embd_k_gqa (hparams.n_embd_k_gqa()), n_embd_head_v (hparams.n_embd_head_v), @@ -7909,6 +8308,7 @@ struct llm_build_context { n_tokens (batch.n_tokens), n_kv (worst_case ? kv_self.size : kv_self.n), n_outputs (worst_case ? n_tokens : lctx.n_outputs), + n_outputs_enc (worst_case ? n_tokens : lctx.embd_enc.size() / hparams.n_embd), kv_head (worst_case ? (kv_self.recurrent ? 0 : kv_self.size - n_tokens) : kv_self.head), n_ctx_orig (cparams.n_ctx_orig_yarn), flash_attn (cparams.flash_attn), @@ -7940,6 +8340,9 @@ struct llm_build_context { lctx.inp_s_copy = nullptr; lctx.inp_s_mask = nullptr; lctx.inp_s_seq = nullptr; + lctx.inp_pos_bucket = nullptr; + lctx.inp_embd_enc = nullptr; + lctx.inp_KQ_mask_cross = nullptr; } void free() { @@ -7959,6 +8362,8 @@ struct llm_build_context { ggml_set_input(lctx.inp_K_shift); for (int il = 0; il < n_layer; ++il) { + const int64_t n_head_kv = hparams.n_head_kv(il); + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(il); struct ggml_tensor * rope_factors = build_rope_factors(il); struct ggml_tensor * tmp = // we rotate only the first n_rot dimensions @@ -8018,6 +8423,9 @@ struct llm_build_context { } for (int il = 0; il < n_layer; ++il) { + const int64_t n_embd_k_gqa = hparams.n_embd_k_gqa(il); + const int64_t n_embd_v_gqa = hparams.n_embd_v_gqa(il); + ggml_tensor * view_k_src = ggml_view_2d(ctx0, kv_self.k_l[il], n_embd_k_gqa, nm, ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa), @@ -8192,6 +8600,53 @@ struct llm_build_context { return gf; } + struct ggml_tensor * llm_build_pos_bucket(bool causal) { + if (causal) { + lctx.inp_pos_bucket = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_kv, n_tokens); + } else { + lctx.inp_pos_bucket = ggml_new_tensor_2d(ctx0, GGML_TYPE_I32, n_tokens, n_tokens); + } + + ggml_set_input(lctx.inp_pos_bucket); + cb(lctx.inp_pos_bucket, "pos_bucket", -1); + + return lctx.inp_pos_bucket; + } + + struct ggml_tensor * llm_build_pos_bias(struct ggml_tensor * pos_bucket, struct ggml_tensor * attn_rel_b) { + struct ggml_tensor * pos_bucket_1d = ggml_view_1d(ctx0, pos_bucket, pos_bucket->ne[0] * pos_bucket->ne[1], 0); + cb(pos_bucket_1d, "pos_bucket_1d", -1); + + struct ggml_tensor * pos_bias = ggml_get_rows(ctx0, attn_rel_b, pos_bucket_1d); + cb(pos_bias, "pos_bias", -1); + + pos_bias = ggml_view_3d(ctx0, pos_bias, pos_bias->ne[0], lctx.inp_pos_bucket->ne[0], lctx.inp_pos_bucket->ne[1], ggml_element_size(pos_bias) * pos_bias->ne[0], ggml_element_size(pos_bias) * pos_bias->ne[0] * lctx.inp_pos_bucket->ne[0], 0); + cb(pos_bias, "pos_bias", -1); + + pos_bias = ggml_permute(ctx0, pos_bias, 2, 0, 1, 3); + cb(pos_bias, "pos_bias", -1); + + pos_bias = ggml_cont(ctx0, pos_bias); + cb(pos_bias, "pos_bias", -1); + + return pos_bias; + } + + struct ggml_tensor * llm_build_inp_embd_enc() { + const int64_t n_embd = hparams.n_embd; + lctx.inp_embd_enc = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_embd, n_outputs_enc); + ggml_set_input(lctx.inp_embd_enc); + cb(lctx.inp_embd_enc, "embd_enc", -1); + return lctx.inp_embd_enc; + } + + struct ggml_tensor * llm_build_inp_KQ_mask_cross() { + lctx.inp_KQ_mask_cross = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_outputs_enc, GGML_PAD(n_tokens, GGML_KQ_MASK_PAD)); + ggml_set_input(lctx.inp_KQ_mask_cross); + cb(lctx.inp_KQ_mask_cross, "KQ_mask_cross", -1); + return lctx.inp_KQ_mask_cross; + } + struct ggml_cgraph * build_llama() { struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false); @@ -11845,6 +12300,131 @@ struct llm_build_context { return gf; } + struct ggml_cgraph * build_openelm() { + struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false); + + const int64_t n_embd_head = hparams.n_embd_head_v; + GGML_ASSERT(n_embd_head == hparams.n_embd_head_k); + + struct ggml_tensor * cur; + struct ggml_tensor * inpL; + inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb); + + // inp_pos - contains the positions + struct ggml_tensor * inp_pos = build_inp_pos(); + + // KQ_mask (mask for 1 head, it will be broadcasted to all heads) + struct ggml_tensor * KQ_mask = build_inp_KQ_mask(); + + for (int il = 0; il < n_layer; ++il) { + const int64_t n_head = hparams.n_head(il); + const int64_t n_head_kv = hparams.n_head_kv(il); + const int64_t n_head_qkv = 2*n_head_kv + n_head; + + cur = inpL; + struct ggml_tensor * residual = cur; + + // norm + cur = llm_build_norm(ctx0, inpL, hparams, + model.layers[il].attn_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "attn_norm", il); + + // self-attention + { + cur = ggml_mul_mat(ctx0, model.layers[il].wqkv, cur); + cb(cur, "wqkv", il); + + cur = ggml_reshape_3d(ctx0, cur, n_embd_head_k, n_head_qkv, n_tokens); + + struct ggml_tensor * Qcur = ggml_cont(ctx0, ggml_view_3d(ctx0, cur, n_embd_head, n_head, n_tokens, cur->nb[1], cur->nb[2], 0)); + cb(Qcur, "Qcur", il); + + struct ggml_tensor * Kcur = ggml_cont(ctx0, ggml_view_3d(ctx0, cur, n_embd_head, n_head_kv, n_tokens, cur->nb[1], cur->nb[2], cur->nb[1]*n_head)); + cb(Kcur, "Kcur", il); + + struct ggml_tensor * Vcur = ggml_cont(ctx0, ggml_view_3d(ctx0, cur, n_embd_head, n_head_kv, n_tokens, cur->nb[1], cur->nb[2], cur->nb[1]*(n_head+n_head_kv))); + cb(Vcur, "Vcur", il); + + Qcur = llm_build_norm(ctx0, Qcur, hparams, + model.layers[il].attn_q_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(Qcur, "Qcur", il); + + Kcur = llm_build_norm(ctx0, Kcur, hparams, + model.layers[il].attn_k_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(Kcur, "Kcur", il); + + Qcur = ggml_rope_ext( + ctx0, Qcur, inp_pos, NULL, n_rot, rope_type, n_ctx_orig, + freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow + ); + cb(Qcur, "Qcur", il); + + Kcur = ggml_rope_ext( + ctx0, Kcur, inp_pos, NULL, n_rot, rope_type, n_ctx_orig, + freq_base, freq_scale, ext_factor, attn_factor, beta_fast, beta_slow + ); + cb(Kcur, "Kcur", il); + + Vcur = ggml_reshape_2d(ctx0, Vcur, n_embd_head * n_head_kv, n_tokens); + cb(Qcur, "Vcur", il); + + cur = llm_build_kv(ctx0, model, hparams, cparams, kv_self, gf, + model.layers[il].wo, NULL, + Kcur, Vcur, Qcur, KQ_mask, n_tokens, kv_head, n_kv, 1.0f/sqrtf(float(n_embd_head)), cb, il); + } + + if (il == n_layer - 1) { + // skip computing output for unused tokens + struct ggml_tensor * inp_out_ids = build_inp_out_ids(); + residual = ggml_get_rows(ctx0, residual, inp_out_ids); + cur = ggml_get_rows(ctx0, cur, inp_out_ids); + } + + struct ggml_tensor * ffn_inp = ggml_add(ctx0, residual, cur); + cb(ffn_inp, "ffn_inp", il); + + // feed-forward network + { + cur = llm_build_norm(ctx0, ffn_inp, hparams, + model.layers[il].ffn_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "ffn_norm", il); + + cur = llm_build_ffn(ctx0, cur, + model.layers[il].ffn_up, NULL, NULL, + model.layers[il].ffn_gate, NULL, NULL, + model.layers[il].ffn_down, NULL, NULL, + NULL, + LLM_FFN_SILU, LLM_FFN_PAR, cb, il); + cb(cur, "ffn_out", il); + } + + cur = ggml_add(ctx0, cur, ffn_inp); + cur = lctx.cvec.apply_to(ctx0, cur, il); + cb(cur, "l_out", il); + + inpL = cur; + } + + cur = inpL; + + // norm + cur = llm_build_norm(ctx0, cur, hparams, + model.output_norm, NULL, + LLM_NORM_RMS, cb, -1); + cb(cur, "result_norm", -1); + + cur = ggml_mul_mat(ctx0, model.output, cur); + cb(cur, "result_output", -1); + + ggml_build_forward_expand(gf, cur); + + return gf; + } + struct ggml_cgraph * build_gptneox() { struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false); @@ -12492,6 +13072,321 @@ struct llm_build_context { return gf; } + struct ggml_cgraph * build_t5() { + struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false); + + // mutable variable, needed during the last layer of the computation to skip unused tokens + int32_t n_tokens = this->n_tokens; + + const int64_t n_embd_head = hparams.n_embd_head_v; + const int64_t n_embd_gqa = hparams.n_embd_v_gqa(); + GGML_ASSERT(n_embd_head == hparams.n_embd_head_k); + + struct ggml_tensor * cur; + struct ggml_tensor * inpL; + + inpL = llm_build_inp_embd(ctx0, lctx, hparams, batch, model.tok_embd, cb); + + if (lctx.is_encoding) { + struct ggml_tensor * pos_bucket_enc = llm_build_pos_bucket(false); + + // KQ_mask (mask for 1 head, it will be broadcasted to all heads) + struct ggml_tensor * KQ_mask_enc = build_inp_KQ_mask(false); + + for (int il = 0; il < n_layer; ++il) { + struct ggml_tensor * inpSA = inpL; + + // norm + cur = llm_build_norm(ctx0, inpL, hparams, + model.layers[il].attn_norm_enc, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "attn_norm", il); + + // self-attention + { + struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq_enc, cur); + cb(Qcur, "Qcur", il); + + struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk_enc, cur); + cb(Kcur, "Kcur", il); + + struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv_enc, cur); + cb(Vcur, "Vcur", il); + + Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens); + Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens); + + struct ggml_tensor * q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3); + struct ggml_tensor * k = ggml_cont(ctx0, ggml_permute(ctx0, Kcur, 0, 2, 1, 3)); + + struct ggml_tensor * kq = ggml_mul_mat(ctx0, k, q); + cb(kq, "kq", il); + + struct ggml_tensor * attn_rel_b = model.layers[il].attn_rel_b_enc ? model.layers[il].attn_rel_b_enc : model.layers[0].attn_rel_b_enc; + struct ggml_tensor * pos_bias = llm_build_pos_bias(pos_bucket_enc, attn_rel_b); + struct ggml_tensor * kq_b = ggml_add(ctx0, kq, pos_bias); + cb(kq_b, "kq_b", il); + + kq = ggml_soft_max_ext(ctx0, kq_b, KQ_mask_enc, 1.0f, hparams.f_max_alibi_bias); + cb(kq, "kq_soft_max_ext", il); + + struct ggml_tensor * v = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_reshape_2d(ctx0, Vcur, n_embd_gqa, n_tokens))); + cb(v, "v", il); + + struct ggml_tensor * kqv = ggml_mul_mat(ctx0, ggml_reshape_3d(ctx0, v, n_tokens, n_embd_head, n_head_kv), kq); + cb(kqv, "kqv", il); + + struct ggml_tensor * kqv_merged = ggml_permute(ctx0, kqv, 0, 2, 1, 3); + cb(kqv_merged, "kqv_merged", il); + + cur = ggml_cont_2d(ctx0, kqv_merged, n_embd_gqa, n_tokens); + cb(cur, "kqv_merged_cont", il); + + ggml_build_forward_expand(gf, cur); + + cur = ggml_mul_mat(ctx0, model.layers[il].wo_enc, cur); + cb(cur, "kqv_out", il); + } + + if (il == n_layer - 1) { + // skip computing output for unused tokens + struct ggml_tensor * inp_out_ids = build_inp_out_ids(); + n_tokens = n_outputs; + cur = ggml_get_rows(ctx0, cur, inp_out_ids); + inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); + } + + struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA); + cb(ffn_inp, "ffn_inp", il); + + // feed-forward network + { + cur = llm_build_norm(ctx0, ffn_inp, hparams, + model.layers[il].ffn_norm_enc, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "ffn_norm", il); + + // T5 uses relu, flan-T5 uses gelu-gated + cur = llm_build_ffn(ctx0, cur, + model.layers[il].ffn_up_enc, NULL, NULL, + model.layers[il].ffn_gate_enc, NULL, NULL, + model.layers[il].ffn_down_enc, NULL, NULL, + NULL, + model.layers[il].ffn_gate_enc ? LLM_FFN_GELU : LLM_FFN_RELU, + model.layers[il].ffn_gate_enc ? LLM_FFN_PAR : LLM_FFN_SEQ, + cb, il); + cb(cur, "ffn_out", il); + } + + cur = ggml_add(ctx0, cur, ffn_inp); + cb(cur, "ffn_out", il); + + ggml_tensor * layer_dir = lctx.cvec.tensor_for(il); + if (layer_dir != nullptr) { + cur = ggml_add(ctx0, cur, layer_dir); + } + cb(cur, "l_out", il); + + // input for next layer + inpL = cur; + } + + cur = inpL; + cb(cur, "result_embd", -1); + + cur = llm_build_norm(ctx0, cur, hparams, + model.output_norm_enc, NULL, + LLM_NORM_RMS, cb, -1); + cb(cur, "result_norm", -1); + } else { + struct ggml_tensor * embd_enc = llm_build_inp_embd_enc(); + struct ggml_tensor * pos_bucket_dec = llm_build_pos_bucket(true); + + struct ggml_tensor * KQ_mask_dec = build_inp_KQ_mask(); + struct ggml_tensor * KQ_mask_cross = llm_build_inp_KQ_mask_cross(); + + for (int il = 0; il < n_layer; ++il) { + struct ggml_tensor * inpSA = inpL; + + // norm + cur = llm_build_norm(ctx0, inpL, hparams, + model.layers[il].attn_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "attn_norm", il); + + // self-attention + { + struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq, cur); + cb(Qcur, "Qcur", il); + + struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk, cur); + cb(Kcur, "Kcur", il); + + struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv, cur); + cb(Vcur, "Vcur", il); + + llm_build_kv_store(ctx0, hparams, cparams, kv_self, gf, Kcur, Vcur, n_tokens, kv_head, cb, il); + + struct ggml_tensor * k = + ggml_view_3d(ctx0, kv_self.k_l[il], + n_embd_head_k, n_kv, n_head_kv, + ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa), + ggml_row_size(kv_self.k_l[il]->type, n_embd_head_k), + 0); + cb(k, "k", il); + + struct ggml_tensor * v = + ggml_view_3d(ctx0, kv_self.v_l[il], + n_kv, n_embd_head_v, n_head_kv, + ggml_element_size(kv_self.v_l[il])*n_ctx, + ggml_element_size(kv_self.v_l[il])*n_ctx*n_embd_head_v, + 0); + cb(v, "v", il); + + Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens); + + struct ggml_tensor * q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3); + + struct ggml_tensor * kq = ggml_mul_mat(ctx0, k, q); + cb(kq, "kq", il); + + struct ggml_tensor * attn_rel_b = model.layers[il].attn_rel_b ? model.layers[il].attn_rel_b : model.layers[0].attn_rel_b; + struct ggml_tensor * pos_bias = llm_build_pos_bias(pos_bucket_dec, attn_rel_b); + struct ggml_tensor * kq_b = ggml_add(ctx0, kq, pos_bias); + cb(kq_b, "kq_b", il); + + kq = ggml_soft_max_ext(ctx0, kq_b, KQ_mask_dec, 1.0f, hparams.f_max_alibi_bias); + cb(kq, "kq_soft_max_ext", il); + + struct ggml_tensor * kqv = ggml_mul_mat(ctx0, v, kq); + cb(kqv, "kqv", il); + + struct ggml_tensor * kqv_merged = ggml_permute(ctx0, kqv, 0, 2, 1, 3); + cb(kqv_merged, "kqv_merged", il); + + cur = ggml_cont_2d(ctx0, kqv_merged, n_embd_gqa, n_tokens); + cb(cur, "kqv_merged_cont", il); + + ggml_build_forward_expand(gf, cur); + + cur = ggml_mul_mat(ctx0, model.layers[il].wo, cur); + cb(cur, "kqv_out", il); + } + + cur = ggml_add(ctx0, cur, inpSA); + cb(cur, "cross_inp", il); + + struct ggml_tensor * inpCA = cur; + + // norm + cur = llm_build_norm(ctx0, cur, hparams, + model.layers[il].attn_norm_cross, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "attn_norm_cross", il); + + // cross-attention + { + struct ggml_tensor * Qcur = ggml_mul_mat(ctx0, model.layers[il].wq_cross, cur); + cb(Qcur, "Qcur", il); + + struct ggml_tensor * Kcur = ggml_mul_mat(ctx0, model.layers[il].wk_cross, embd_enc); + cb(Kcur, "Kcur", il); + + struct ggml_tensor * Vcur = ggml_mul_mat(ctx0, model.layers[il].wv_cross, embd_enc); + cb(Vcur, "Vcur", il); + + Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head, n_tokens); + Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_outputs_enc); + + struct ggml_tensor * q = ggml_permute(ctx0, Qcur, 0, 2, 1, 3); + struct ggml_tensor * k = ggml_cont(ctx0, ggml_permute(ctx0, Kcur, 0, 2, 1, 3)); + + struct ggml_tensor * kq = ggml_mul_mat(ctx0, k, q); + cb(kq, "kq", il); + + kq = ggml_soft_max_ext(ctx0, kq, KQ_mask_cross, 1.0f, hparams.f_max_alibi_bias); + cb(kq, "kq_soft_max_ext", il); + + struct ggml_tensor * v = ggml_cont(ctx0, ggml_transpose(ctx0, ggml_reshape_2d(ctx0, Vcur, n_embd_gqa, n_outputs_enc))); + cb(v, "v", il); + + struct ggml_tensor * kqv = ggml_mul_mat(ctx0, ggml_reshape_3d(ctx0, v, n_outputs_enc, n_embd_head, n_head_kv), kq); + cb(kqv, "kqv", il); + + struct ggml_tensor * kqv_merged = ggml_permute(ctx0, kqv, 0, 2, 1, 3); + cb(kqv_merged, "kqv_merged", il); + + cur = ggml_cont_2d(ctx0, kqv_merged, n_embd_gqa, n_tokens); + cb(cur, "kqv_merged_cont", il); + + ggml_build_forward_expand(gf, cur); + + cur = ggml_mul_mat(ctx0, model.layers[il].wo_cross, cur); + cb(cur, "kqv_out", il); + } + + if (il == n_layer - 1) { + // skip computing output for unused tokens + struct ggml_tensor * inp_out_ids = build_inp_out_ids(); + n_tokens = n_outputs; + cur = ggml_get_rows(ctx0, cur, inp_out_ids); + inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids); + inpCA = ggml_get_rows(ctx0, inpCA, inp_out_ids); + } + + struct ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpCA); + cb(ffn_inp, "ffn_inp", il); + + // feed-forward network + { + cur = llm_build_norm(ctx0, ffn_inp, hparams, + model.layers[il].ffn_norm, NULL, + LLM_NORM_RMS, cb, il); + cb(cur, "ffn_norm", il); + + // T5 uses relu, flan-T5 uses gelu-gated + cur = llm_build_ffn(ctx0, cur, + model.layers[il].ffn_up, NULL, NULL, + model.layers[il].ffn_gate, NULL, NULL, + model.layers[il].ffn_down, NULL, NULL, + NULL, + model.layers[il].ffn_gate_enc ? LLM_FFN_GELU : LLM_FFN_RELU, + model.layers[il].ffn_gate_enc ? LLM_FFN_PAR : LLM_FFN_SEQ, + cb, il); + cb(cur, "ffn_out", il); + } + + cur = ggml_add(ctx0, cur, ffn_inp); + cb(cur, "ffn_out", il); + + ggml_tensor * layer_dir = lctx.cvec.tensor_for(il); + if (layer_dir != nullptr) { + cur = ggml_add(ctx0, cur, layer_dir); + } + cb(cur, "l_out", il); + + // input for next layer + inpL = cur; + } + + cur = inpL; + cb(cur, "result_embd", -1); + + cur = llm_build_norm(ctx0, cur, hparams, + model.output_norm, NULL, + LLM_NORM_RMS, cb, -1); + cb(cur, "result_norm", -1); + + // lm_head + cur = ggml_mul_mat(ctx0, model.output, cur); + cb(cur, "result_output", -1); + } + + ggml_build_forward_expand(gf, cur); + + return gf; + } + struct ggml_cgraph * build_jais() { struct ggml_cgraph * gf = ggml_new_graph_custom(ctx0, LLAMA_MAX_NODES, false); @@ -12912,6 +13807,10 @@ static struct ggml_cgraph * llama_build_graph( { result = llm.build_olmo(); } break; + case LLM_ARCH_OPENELM: + { + result = llm.build_openelm(); + } break; case LLM_ARCH_GPTNEOX: { result = llm.build_gptneox(); @@ -12932,6 +13831,10 @@ static struct ggml_cgraph * llama_build_graph( { result = llm.build_bitnet(); } break; + case LLM_ARCH_T5: + { + result = llm.build_t5(); + } break; case LLM_ARCH_JAIS: { result = llm.build_jais(); @@ -12974,6 +13877,30 @@ static void llama_set_s_copy(llama_context & lctx) { } } +static int32_t llama_relative_position_bucket(llama_pos x, llama_pos y, uint64_t n_buckets, bool bidirectional) { + // TODO move to hparams if a T5 variant appears that uses a different value + const int64_t max_distance = 128; + + if (bidirectional) { + n_buckets >>= 1; + } + + const int64_t max_exact = n_buckets >> 1; + + int32_t relative_position = x - y; + int32_t relative_bucket = 0; + if (bidirectional) { + relative_bucket += (relative_position > 0) * n_buckets; + relative_position = abs(relative_position); + } else { + relative_position = -std::min(relative_position, 0); + } + int32_t relative_position_if_large = floorf(max_exact + logf(1.0 * relative_position / max_exact) * (n_buckets - max_exact) / log(1.0 * max_distance / max_exact)); + relative_position_if_large = std::min(relative_position_if_large, n_buckets - 1); + relative_bucket += (relative_position < max_exact ? relative_position : relative_position_if_large); + return relative_bucket; +} + static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { // // set input data @@ -13039,7 +13966,7 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { if (lctx.inp_KQ_mask) { // NOTE: hparams.causal_attn indicates the model is capable of generation and uses the kv cache. - if (cparams.causal_attn) { + if (cparams.causal_attn && !lctx.is_encoding) { const int64_t n_kv = kv_self.n; const int64_t n_tokens = batch.n_tokens; @@ -13092,7 +14019,7 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { } else { // when using kv cache, the mask needs to match the kv cache size const int64_t n_tokens = batch.n_tokens; - const int64_t n_stride = hparams.causal_attn ? kv_self.n : n_tokens; + const int64_t n_stride = hparams.causal_attn && !lctx.is_encoding ? kv_self.n : n_tokens; GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_KQ_mask->buffer)); @@ -13126,7 +14053,7 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { } } - if (cparams.pooling_type == LLAMA_POOLING_TYPE_MEAN) { + if (cparams.embeddings && cparams.pooling_type == LLAMA_POOLING_TYPE_MEAN) { const int64_t n_tokens = batch.n_tokens; GGML_ASSERT(lctx.inp_mean); @@ -13158,7 +14085,7 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { } } - if (cparams.pooling_type == LLAMA_POOLING_TYPE_CLS) { + if (cparams.embeddings && cparams.pooling_type == LLAMA_POOLING_TYPE_CLS) { const int64_t n_tokens = batch.n_tokens; GGML_ASSERT(lctx.inp_cls); @@ -13179,7 +14106,7 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { } } - if (cparams.pooling_type == LLAMA_POOLING_TYPE_LAST) { + if (cparams.embeddings && cparams.pooling_type == LLAMA_POOLING_TYPE_LAST) { const int64_t n_tokens = batch.n_tokens; GGML_ASSERT(lctx.inp_cls); @@ -13256,6 +14183,70 @@ static void llama_set_inputs(llama_context & lctx, const llama_batch & batch) { } } } + + if (lctx.inp_pos_bucket) { + const int64_t n_tokens = batch.n_tokens; + + GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_pos_bucket->buffer)); + + int32_t * data = (int32_t *) lctx.inp_pos_bucket->data; + + if (!lctx.is_encoding) { + const int64_t n_kv = kv_self.n; + for (int h = 0; h < 1; ++h) { + for (int j = 0; j < n_tokens; ++j) { + for (int i = 0; i < n_kv; ++i) { + data[h*(n_kv*n_tokens) + j*n_kv + i] = llama_relative_position_bucket(lctx.kv_self.cells[i].pos, batch.pos[j], hparams.n_rel_attn_bkts, lctx.is_encoding); + } + } + } + } else { + for (int h = 0; h < 1; ++h) { + for (int j = 0; j < n_tokens; ++j) { + for (int i = 0; i < n_tokens; ++i) { + data[h*(n_tokens*n_tokens) + j*n_tokens + i] = llama_relative_position_bucket(batch.pos[i], batch.pos[j], hparams.n_rel_attn_bkts, lctx.is_encoding); + } + } + } + } + } + + if (!lctx.is_encoding && lctx.inp_embd_enc) { + assert(lctx.inp_embd_enc->type == GGML_TYPE_F32); + assert((size_t) ggml_nelements(lctx.inp_embd_enc) == lctx.embd_enc.size()); + + ggml_backend_tensor_set(lctx.inp_embd_enc, lctx.embd_enc.data(), 0, ggml_nbytes(lctx.inp_embd_enc)); + } + + if (!lctx.is_encoding && lctx.inp_KQ_mask_cross) { + const int64_t n_output_enc = lctx.embd_enc.size() / hparams.n_embd; + const int64_t n_tokens = batch.n_tokens; + + GGML_ASSERT(ggml_backend_buffer_is_host(lctx.inp_KQ_mask_cross->buffer)); + + float * data = (float *) lctx.inp_KQ_mask_cross->data; + + for (int h = 0; h < 1; ++h) { + for (int j = 0; j < n_tokens; ++j) { + for (int i = 0; i < n_output_enc; ++i) { + float f = -INFINITY; + for (int s = 0; s < batch.n_seq_id[j]; ++s) { + const llama_seq_id seq_id = batch.seq_id[j][s]; + if (lctx.seq_ids_enc[i].find(seq_id) != lctx.seq_ids_enc[i].end()) { + f = 0.0f; + } + } + data[h*(n_output_enc*n_tokens) + j*n_output_enc + i] = f; + } + } + + for (int i = n_tokens; i < GGML_PAD(n_tokens, GGML_KQ_MASK_PAD); ++i) { + for (int j = 0; j < n_output_enc; ++j) { + data[h*(n_output_enc*n_tokens) + i*n_output_enc + j] = -INFINITY; + } + } + } + } } // Make sure enough space is available for outputs. @@ -13272,7 +14263,7 @@ static size_t llama_output_reserve(llama_context & lctx, size_t n_outputs) { // TODO: use a per-batch flag for logits presence instead const bool has_logits = !cparams.embeddings; - const bool has_embd = cparams.embeddings && (cparams.pooling_type == LLAMA_POOLING_TYPE_NONE); + const bool has_embd = lctx.is_encoding || (cparams.embeddings && (cparams.pooling_type == LLAMA_POOLING_TYPE_NONE)); const size_t logits_size = has_logits ? n_vocab*n_outputs_max : 0; const size_t embd_size = has_embd ? n_embd*n_outputs_max : 0; @@ -13364,6 +14355,7 @@ static int llama_decode_internal( llama_context & lctx, llama_batch batch_all) { // TODO: rename back to batch + lctx.is_encoding = false; const uint32_t n_tokens_all = batch_all.n_tokens; if (n_tokens_all == 0) { @@ -13396,19 +14388,21 @@ static int llama_decode_internal( const auto n_ubatch = cparams.n_ubatch; + // TODO: simplify or deprecate std::vector pos; std::vector n_seq_id; std::vector seq_id_arr; std::vector> seq_id; + // this indicates we are doing pooled embedding, so we ignore batch.logits and output all tokens + const bool embd_pooled = cparams.embeddings && cparams.pooling_type != LLAMA_POOLING_TYPE_NONE; + // count outputs - if (cparams.embeddings && cparams.pooling_type != LLAMA_POOLING_TYPE_NONE) { - n_outputs = n_tokens_all; - } else if (batch_all.logits) { + if (batch_all.logits && !embd_pooled) { for (uint32_t i = 0; i < n_tokens_all; ++i) { n_outputs += batch_all.logits[i] != 0; } - } else if (lctx.logits_all) { + } else if (lctx.logits_all || embd_pooled) { n_outputs = n_tokens_all; } else { // keep last output only @@ -13454,7 +14448,7 @@ static int llama_decode_internal( { int32_t n_outputs_new = 0; - if (u_batch.logits) { + if (u_batch.logits && !embd_pooled) { for (uint32_t i = 0; i < n_tokens; i++) { n_outputs_new += u_batch.logits[i] != 0; } @@ -13659,6 +14653,138 @@ static int llama_decode_internal( return 0; } +// encode a batch of tokens by evaluating the encoder part of the transformer +// +// - lctx: llama context +// - batch: batch to evaluate +// +// return 0 on success +// return positive int on warning +// return negative int on error +// +static int llama_encode_internal( + llama_context & lctx, + llama_batch batch) { + + lctx.is_encoding = true; + + const uint32_t n_tokens = batch.n_tokens; + + if (n_tokens == 0) { + LLAMA_LOG_ERROR("%s: n_tokens == 0", __func__); + return -1; + } + + const auto & model = lctx.model; + const auto & hparams = model.hparams; + const auto & cparams = lctx.cparams; + + GGML_ASSERT((!batch.token && batch.embd) || (batch.token && !batch.embd)); // NOLINT + + // micro-batching is not possible for non-causal encoding, so we process the batch in a single shot + GGML_ASSERT(cparams.n_ubatch >= n_tokens && "encoder requires n_ubatch >= n_tokens"); + + if (lctx.t_compute_start_us == 0) { + lctx.t_compute_start_us = ggml_time_us(); + } + + lctx.n_queued_tokens += n_tokens; + + const int64_t n_embd = hparams.n_embd; + + // TODO: simplify or deprecate + std::vector pos; + std::vector n_seq_id; + std::vector seq_id_arr; + std::vector> seq_id; + + // reserve output buffer + if (llama_output_reserve(lctx, n_tokens) < n_tokens) { + LLAMA_LOG_ERROR("%s: could not reserve space for batch with %u outputs\n", __func__, n_tokens); + return -2; + }; + + for (uint32_t i = 0; i < n_tokens; ++i) { + lctx.output_ids[i] = i; + } + + lctx.inp_embd_enc = NULL; + lctx.n_outputs = n_tokens; + + const int n_threads = n_tokens == 1 ? cparams.n_threads : cparams.n_threads_batch; + GGML_ASSERT(n_threads > 0); + + // helpers for smoother batch API transition + // after deprecating the llama_eval calls, these will be removed + if (batch.pos == nullptr) { + pos.resize(n_tokens); + for (uint32_t i = 0; i < n_tokens; i++) { + pos[i] = batch.all_pos_0 + i*batch.all_pos_1; + } + + batch.pos = pos.data(); + } + + if (batch.seq_id == nullptr) { + n_seq_id.resize(n_tokens); + seq_id.resize(n_tokens); + seq_id_arr.resize(n_tokens); + for (uint32_t i = 0; i < n_tokens; i++) { + n_seq_id[i] = 1; + seq_id[i].resize(1); + seq_id[i][0] = batch.all_seq_id; + seq_id_arr[i] = seq_id[i].data(); + } + + batch.n_seq_id = n_seq_id.data(); + batch.seq_id = seq_id_arr.data(); + } + + ggml_backend_sched_reset(lctx.sched); + ggml_backend_sched_set_eval_callback(lctx.sched, lctx.cparams.cb_eval, lctx.cparams.cb_eval_user_data); + + ggml_cgraph * gf = llama_build_graph(lctx, batch, false); + + // the output embeddings after the final encoder normalization + struct ggml_tensor * embd = gf->nodes[gf->n_nodes - 1]; + + GGML_ASSERT(strcmp(embd->name, "result_norm") == 0); + + ggml_backend_sched_alloc_graph(lctx.sched, gf); + + llama_set_inputs(lctx, batch); + + llama_graph_compute(lctx, gf, n_threads); + + // extract embeddings + if (embd) { + ggml_backend_t backend_embd = ggml_backend_sched_get_tensor_backend(lctx.sched, embd); + GGML_ASSERT(backend_embd != nullptr); + + // extract token embeddings + GGML_ASSERT(lctx.embd != nullptr); + + lctx.embd_enc.resize(n_tokens*n_embd); + float * embd_out = lctx.embd_enc.data(); + + ggml_backend_tensor_get_async(backend_embd, embd, embd_out, 0, n_tokens*n_embd*sizeof(float)); + + // remember the sequence ids used during the encoding - needed for cross attention later + lctx.seq_ids_enc.resize(n_tokens); + for (uint32_t i = 0; i < n_tokens; i++) { + for (int s = 0; s < batch.n_seq_id[i]; s++) { + llama_seq_id seq_id = batch.seq_id[i][s]; + lctx.seq_ids_enc[i].insert(seq_id); + } + } + } + + // Reset state for the next token before backend sync, to allow the CPU activities in the reset to + // overlap with device computation. + ggml_backend_sched_reset(lctx.sched); + + return 0; +} // find holes from the beginning of the KV cache and fill them by moving data from the end of the cache static void llama_kv_cache_defrag_internal(struct llama_context & lctx) { @@ -14736,11 +15862,14 @@ struct llm_tokenizer_ugm { std::string normalized; normalize(text, &normalized); size_t input_len = normalized.size(); + if (input_len == 0) { + return; + } // initialize score_sum to -FLT_MAX so it will be always lower than sums of token scores - std::vector tokenization_results(input_len + 1, {0, 0, -FLT_MAX}); + std::vector tokenization_results(input_len + 1, {vocab.special_unk_id, 0, -FLT_MAX}); // at the beginning tokenization score is zero - tokenization_results[0] = { 0, 0, 0 }; + tokenization_results[0] = { vocab.special_unk_id, 0, 0 }; for (size_t input_offset = 0; input_offset < input_len;) { size_t prefix_offset = input_offset; @@ -14760,7 +15889,7 @@ struct llm_tokenizer_ugm { single_codepoint_token_found = true; } llama_token token_id = node->value; - const auto &token_data = vocab.id_to_token[token_id]; + const auto & token_data = vocab.id_to_token[token_id]; // we set the user-defined token scores to 0 to make them more likely to be selected // (normal token scores are log probabilities, so they are negative) @@ -14970,7 +16099,7 @@ private: size_t prefix_offset = input_offset; unicode_cpt_from_utf8(input, prefix_offset); return { &input[input_offset], prefix_offset - input_offset, prefix_offset - input_offset }; - } catch(std::invalid_argument & ex) { + } catch (std::invalid_argument & /*ex*/) { // if no, consume 1 byte and return U+FFFD - REPLACEMENT CHARACTER return { "\xEF\xBF\xBD", 3, 1 }; } @@ -15170,7 +16299,8 @@ static std::vector llama_tokenize_internal(const llama_vocab & // tokenizer.encode('', add_special_tokens=True) returns [1] // tokenizer.encode('', add_special_tokens=False) returns [] - bool is_prev_special = false; + bool is_prev_special = true; // prefix with space if first token + if (add_special && vocab.tokenizer_add_bos) { GGML_ASSERT(vocab.special_bos_id != -1); output.push_back(vocab.special_bos_id); @@ -15181,10 +16311,9 @@ static std::vector llama_tokenize_internal(const llama_vocab & if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_RAW_TEXT) { auto raw_text = fragment.raw_text.substr(fragment.offset, fragment.length); - if (vocab.tokenizer_add_space_prefix) { - if (!output.size() || is_prev_special) { // prefix with space if first token - raw_text = " " + raw_text; - } + // prefix with space if previous is special + if (vocab.tokenizer_add_space_prefix && is_prev_special) { + raw_text = " " + raw_text; } #ifdef PRETOKENIZERDEBUG @@ -15193,6 +16322,7 @@ static std::vector llama_tokenize_internal(const llama_vocab & llm_tokenizer_spm tokenizer(vocab); llama_escape_whitespace(raw_text); tokenizer.tokenize(raw_text, output); + is_prev_special = false; } else { // if (fragment.type == FRAGMENT_BUFFER_VARIANT_TYPE_TOKEN) output.push_back(fragment.token); is_prev_special = true; @@ -16576,8 +17706,8 @@ static ggml_type llama_tensor_get_type(quantize_state_internal & qs, ggml_type n const llm_arch arch = qs.model.arch; const auto tn = LLM_TN(arch); - auto use_more_bits = [](int i_layer, int num_layers) -> bool { - return i_layer < num_layers/8 || i_layer >= 7*num_layers/8 || (i_layer - num_layers/8)%3 == 2; + auto use_more_bits = [](int i_layer, int n_layers) -> bool { + return i_layer < n_layers/8 || i_layer >= 7*n_layers/8 || (i_layer - n_layers/8)%3 == 2; }; const int n_expert = std::max(1, (int)qs.model.hparams.n_expert); auto layer_info = [n_expert] (int i_layer, int n_layer, const char * name) { @@ -17041,10 +18171,11 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s // sanity checks // - // - qs.n_attention_wv == 0 for Mamba models - // - qs.n_attention_wv == model.hparams.n_layer for Transformer models + // - qs.n_attention_wv == 0 for Mamba models + // - qs.n_attention_wv == model.hparams.n_layer for Transformer models + // - qs.n_attention_wv == 3 * model.hparams.n_layer for Encoder-Decoder models // - GGML_ASSERT((qs.n_attention_wv == 0 || qs.n_attention_wv == (int)model.hparams.n_layer) && "n_attention_wv is unexpected"); + GGML_ASSERT((qs.n_attention_wv == 0 || qs.n_attention_wv == (int)model.hparams.n_layer || qs.n_attention_wv == 3 * (int)model.hparams.n_layer) && "n_attention_wv is unexpected"); size_t total_size_org = 0; size_t total_size_new = 0; @@ -17169,6 +18300,9 @@ static void llama_model_quantize_internal(const std::string & fname_inp, const s quantize &= name.find("ssm_x.weight") == std::string::npos; quantize &= name.find("ssm_dt.weight") == std::string::npos; + // do not quantize relative position bias (T5) + quantize &= name.find("attn_rel_b.weight") == std::string::npos; + enum ggml_type new_type; void * new_data; size_t new_size; @@ -17617,6 +18751,7 @@ struct llama_context_params llama_context_default_params() { /*.n_threads_batch =*/ GGML_DEFAULT_N_THREADS, /*.rope_scaling_type =*/ LLAMA_ROPE_SCALING_TYPE_UNSPECIFIED, /*.pooling_type =*/ LLAMA_POOLING_TYPE_UNSPECIFIED, + /*.attention_type =*/ LLAMA_ATTENTION_TYPE_UNSPECIFIED, /*.rope_freq_base =*/ 0.0f, /*.rope_freq_scale =*/ 0.0f, /*.yarn_ext_factor =*/ -1.0f, @@ -17869,7 +19004,6 @@ struct llama_context * llama_new_context_with_model( } cparams.yarn_attn_factor *= hparams.rope_attn_factor; - cparams.causal_attn = hparams.causal_attn; if (cparams.pooling_type == LLAMA_POOLING_TYPE_UNSPECIFIED) { if (hparams.pooling_type == LLAMA_POOLING_TYPE_UNSPECIFIED) { @@ -17879,6 +19013,12 @@ struct llama_context * llama_new_context_with_model( } } + if (params.attention_type == LLAMA_ATTENTION_TYPE_UNSPECIFIED) { + cparams.causal_attn = hparams.causal_attn; + } else { + cparams.causal_attn = params.attention_type == LLAMA_ATTENTION_TYPE_CAUSAL; + } + if (params.seed == LLAMA_DEFAULT_SEED) { params.seed = time(NULL); } @@ -18217,6 +19357,7 @@ enum llama_rope_type llama_rope_type(const struct llama_model * model) { case LLM_ARCH_GEMMA: case LLM_ARCH_GEMMA2: case LLM_ARCH_STARCODER2: + case LLM_ARCH_OPENELM: case LLM_ARCH_GPTNEOX: return LLAMA_ROPE_TYPE_NEOX; @@ -18326,6 +19467,17 @@ struct ggml_tensor * llama_get_model_tensor(struct llama_model * model, const ch return it->second; } +bool llama_model_has_encoder(const struct llama_model * model) { + switch (model->arch) { + case LLM_ARCH_T5: return true; + default: return false; + } +} + +llama_token llama_model_decoder_start_token(const struct llama_model * model) { + return model->hparams.dec_start_token_id; +} + uint32_t llama_model_quantize( const char * fname_inp, const char * fname_out, @@ -18799,8 +19951,6 @@ static void llama_state_get_data_internal(struct llama_context * ctx, llama_data const auto & hparams = ctx->model.hparams; const uint32_t n_layer = hparams.n_layer; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); // NOTE: kv_size and kv_buf_size are mostly used for sanity checks const uint32_t kv_head = llama_kv_cache_cell_max(kv_self); @@ -18820,6 +19970,9 @@ static void llama_state_get_data_internal(struct llama_context * ctx, llama_data std::vector tmp_buf; for (int il = 0; il < (int) n_layer; ++il) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(); + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + const size_t k_size = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*kv_head); tmp_buf.resize(k_size); @@ -18952,8 +20105,6 @@ size_t llama_state_set_data(struct llama_context * ctx, const uint8_t * src) { const auto & hparams = ctx->model.hparams; const uint32_t n_layer = hparams.n_layer; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); size_t kv_buf_size; uint32_t kv_head; @@ -18985,6 +20136,9 @@ size_t llama_state_set_data(struct llama_context * ctx, const uint8_t * src) { GGML_ASSERT(kv_self.total_size() >= kv_buf_size); for (int il = 0; il < (int) n_layer; ++il) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(); + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + const size_t k_size = ggml_row_size(kv_self.k_l[il]->type, n_embd_k_gqa*kv_head); ggml_backend_tensor_set(kv_self.k_l[il], inp, 0, k_size); @@ -19147,8 +20301,6 @@ size_t llama_state_seq_get_size(struct llama_context* ctx, llama_seq_id seq_id) const auto & hparams = ctx->model.hparams; const uint32_t n_layer = hparams.n_layer; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); for (uint32_t i = 0; i < kv_self.size; ++i) { const auto & cell = kv_self.cells[i]; @@ -19159,6 +20311,9 @@ size_t llama_state_seq_get_size(struct llama_context* ctx, llama_seq_id seq_id) } for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(); + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + // types of keys and values s_cell_data_size += sizeof(int32_t) * 2; // k_size_row and v_size_el values of layer @@ -19233,14 +20388,15 @@ static size_t llama_state_seq_get_data_internal(struct llama_context * ctx, llam const auto & hparams = ctx->model.hparams; const uint32_t n_layer = hparams.n_layer; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); // Write the layer count data_ctx.write(&n_layer, sizeof(n_layer)); - // Write n_embd_v_gqa - data_ctx.write(&n_embd_v_gqa, sizeof(n_embd_v_gqa)); + // Write n_embd_v_gqa (reference value) + { + const uint32_t n_embd_v_gqa_ref = hparams.n_embd_v_gqa() + hparams.n_embd_k_s(); + data_ctx.write(&n_embd_v_gqa_ref, sizeof(n_embd_v_gqa_ref)); + } // Iterate the ranges and write all the pos (this is the token position in the prompt) for (const auto & range : cell_ranges) { @@ -19254,6 +20410,8 @@ static size_t llama_state_seq_get_data_internal(struct llama_context * ctx, llam // Get whole range at a time std::vector tmp_buf; for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(); + // Write key type const int32_t k_type_i = (int32_t)kv_self.k_l[il]->type; data_ctx.write(&k_type_i, sizeof(k_type_i)); @@ -19274,6 +20432,8 @@ static size_t llama_state_seq_get_data_internal(struct llama_context * ctx, llam // TODO: simplify, reduce copy-paste if (!kv_self.v_trans) { for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + // Write value type const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type; data_ctx.write(&v_type_i, sizeof(v_type_i)); @@ -19294,6 +20454,8 @@ static size_t llama_state_seq_get_data_internal(struct llama_context * ctx, llam // For the values, they are transposed, so we also need the element size and get the element ranges from each row const uint32_t kv_size = kv_self.size; for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + // Write value type const int32_t v_type_i = (int32_t)kv_self.v_l[il]->type; data_ctx.write(&v_type_i, sizeof(v_type_i)); @@ -19362,14 +20524,14 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src, // Sanity check model compatibility const auto & hparams = ctx->model.hparams; const uint32_t n_layer = hparams.n_layer; - const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa() + hparams.n_embd_k_s(); - const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa() + hparams.n_embd_v_s(); + if (n_layer != n_layer_ref) { LLAMA_LOG_ERROR("%s: mismatched n_layer (%d != %d)\n", __func__, n_layer, n_layer_ref); return 0; } - if (n_embd_v_gqa != n_embd_v_gqa_ref) { - LLAMA_LOG_ERROR("%s: mismatched n_embd_v_gqa (%d != %d)\n", __func__, n_embd_v_gqa, n_embd_v_gqa_ref); + + if (hparams.n_embd_v_gqa() != n_embd_v_gqa_ref) { + LLAMA_LOG_ERROR("%s: mismatched n_embd_v_gqa (%d != %d)\n", __func__, hparams.n_embd_v_gqa(), n_embd_v_gqa_ref); return 0; } @@ -19409,6 +20571,8 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src, // For each layer, read the keys for each cell, one row is one cell, read as one contiguous blo for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_k_gqa = hparams.n_embd_k_gqa(il) + hparams.n_embd_k_s(); + // Read type of key int32_t k_type_i_ref; memcpy(&k_type_i_ref, inp, sizeof(k_type_i_ref)); @@ -19441,6 +20605,8 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src, // TODO: simplify, reduce copy-paste if (!kv_self.v_trans) { for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + // Read type of value int32_t v_type_i_ref; memcpy(&v_type_i_ref, inp, sizeof(v_type_i_ref)); @@ -19472,6 +20638,8 @@ size_t llama_state_seq_set_data(struct llama_context * ctx, const uint8_t * src, } else { // For each layer, read the values for each cell (transposed) for (int il = 0; il < (int)n_layer; ++il) { + const uint32_t n_embd_v_gqa = hparams.n_embd_v_gqa(il) + hparams.n_embd_v_s(); + // Read type of value int32_t v_type_i_ref; memcpy(&v_type_i_ref, inp, sizeof(v_type_i_ref)); @@ -19672,6 +20840,17 @@ void llama_batch_free(struct llama_batch batch) { if (batch.logits) free(batch.logits); } +int32_t llama_encode( + struct llama_context * ctx, + struct llama_batch batch) { + const int ret = llama_encode_internal(*ctx, batch); + if (ret < 0) { + LLAMA_LOG_ERROR("%s: failed to encode, ret = %d\n", __func__, ret); + } + + return ret; +} + int32_t llama_decode( struct llama_context * ctx, struct llama_batch batch) { @@ -19912,7 +21091,7 @@ static std::string llama_decode_text(const std::string & text) { const auto utf8 = unicode_cpt_to_utf8(cpt); try { decoded_text += unicode_utf8_to_byte(utf8); - } catch (const std::out_of_range & e) { + } catch (const std::out_of_range & /*e*/) { decoded_text += "[UNK_BYTE_0x"; for (const auto c : utf8) { decoded_text += format("%02x", (uint8_t) c); @@ -19925,85 +21104,66 @@ static std::string llama_decode_text(const std::string & text) { } // does not write null-terminator to buf -int32_t llama_token_to_piece(const struct llama_model * model, llama_token token, char * buf, int32_t length, bool special) { +int32_t llama_token_to_piece(const struct llama_model * model, llama_token token, char * buf, int32_t length, int32_t lstrip, bool special) { // ref: https://github.com/ggerganov/llama.cpp/pull/7587#discussion_r1620983843 - if (!special && llama_is_control_token(model->vocab, token)) { + static const int attr_special = LLAMA_TOKEN_ATTR_UNKNOWN | LLAMA_TOKEN_ATTR_CONTROL; + const llama_token_attr attr = llama_token_get_attr(model, token); + if (!special && (attr & attr_special)) { return 0; } + // copy piece chars to output text buffer + // skip up to 'lstrip' leading spaces before copying + auto _try_copy = [=] (const char * token, size_t size) -> int32_t { + for (int32_t i = 0; i < lstrip && size && *token == ' '; ++i) { + token++; + size--; + } + if (length < (int32_t)size) { + return (int32_t) -size; + } + memcpy(buf, token, size); + return (int32_t) size; + }; + // if we have a cache - use it { const auto & cache = model->vocab.cache_token_to_piece; if (!cache.empty()) { - const auto & res = cache.at(token); - if (length < (int) res.size()) { - return -(int) res.size(); - } - memcpy(buf, res.c_str(), res.size()); - return res.size(); + const auto & result = cache.at(token); + return _try_copy(result.data(), result.size()); } } if (0 <= token && token < llama_n_vocab(model)) { + const std::string & token_text = model->vocab.id_to_token[token].text; switch (llama_vocab_get_type(model->vocab)) { case LLAMA_VOCAB_TYPE_WPM: case LLAMA_VOCAB_TYPE_SPM: case LLAMA_VOCAB_TYPE_UGM: { // NOTE: we accept all unsupported token types, // suppressing them like CONTROL tokens. - if (llama_is_normal_token(model->vocab, token)) { - std::string result = model->vocab.id_to_token[token].text; + if (attr & (attr_special | LLAMA_TOKEN_ATTR_USER_DEFINED)) { + return _try_copy(token_text.data(), token_text.size()); + } else if (attr & LLAMA_TOKEN_ATTR_NORMAL) { + std::string result = token_text; llama_unescape_whitespace(result); - if (length < (int) result.length()) { - return -(int) result.length(); - } - memcpy(buf, result.c_str(), result.length()); - return result.length(); - } else if ( - (llama_is_user_defined_token(model->vocab, token)) || - (llama_is_control_token (model->vocab, token) && special)) { - std::string result = model->vocab.id_to_token[token].text; - if (length < (int) result.length()) { - return -(int) result.length(); - } - memcpy(buf, result.c_str(), result.length()); - return result.length(); - } else if (llama_is_unknown_token(model->vocab, token)) { // NOLINT - if (length < 3) { - return -3; - } - memcpy(buf, "\xe2\x96\x85", 3); - return 3; - } else if (llama_is_byte_token(model->vocab, token)) { - if (length < 1) { - return -1; - } - buf[0] = llama_token_to_byte(model->vocab, token); - return 1; + return _try_copy(result.data(), result.size()); + } else if (attr & LLAMA_TOKEN_ATTR_BYTE) { + char byte = (char) llama_token_to_byte(model->vocab, token); + return _try_copy((char*) &byte, 1); } break; } case LLAMA_VOCAB_TYPE_BPE: { // NOTE: we accept all unsupported token types, // suppressing them like CONTROL tokens. - if (llama_is_normal_token(model->vocab, token)) { - std::string result = model->vocab.id_to_token[token].text; - result = llama_decode_text(result); - if (length < (int) result.length()) { - return -(int) result.length(); - } - memcpy(buf, result.c_str(), result.length()); - return result.length(); - } else if ( - (llama_is_user_defined_token(model->vocab, token)) || - (llama_is_control_token (model->vocab, token) && special)) { - std::string result = model->vocab.id_to_token[token].text; - if (length < (int) result.length()) { - return -(int) result.length(); - } - memcpy(buf, result.c_str(), result.length()); - return result.length(); + if (attr & (attr_special | LLAMA_TOKEN_ATTR_USER_DEFINED)) { + return _try_copy(token_text.data(), token_text.size()); + } else if (attr & LLAMA_TOKEN_ATTR_NORMAL) { + std::string result = llama_decode_text(token_text); + return _try_copy(result.data(), result.size()); } break; } @@ -20014,6 +21174,113 @@ int32_t llama_token_to_piece(const struct llama_model * model, llama_token token return 0; } +int32_t llama_detokenize( + const struct llama_model * model, + const llama_token * tokens, + int32_t n_tokens, + char * text, + int32_t text_len_max, + bool remove_special, + bool unparse_special) { + int32_t avail = text_len_max; + int32_t total = 0; + + // remove the leading space + bool remove_space = model->vocab.tokenizer_add_space_prefix; + + if (remove_special && model->vocab.tokenizer_add_bos) { + if (n_tokens > 0 && tokens[0] == model->vocab.special_bos_id) { + remove_space = false; + n_tokens--; + tokens++; + } + } + + if (remove_special && model->vocab.tokenizer_add_eos) { + if (n_tokens > 0 && tokens[n_tokens-1] == model->vocab.special_eos_id) { + n_tokens--; + } + } + + for (int32_t i = 0; i < n_tokens; ++i) { + GGML_ASSERT(avail >= 0); + int32_t n_chars = llama_token_to_piece(model, tokens[i], text, avail, remove_space, unparse_special); + remove_space = false; + if (n_chars < 0) { + avail = 0; + total -= n_chars; + } else if (n_chars > 0) { + avail -= n_chars; + text += n_chars; + total += n_chars; + } + } + + if (total > text_len_max) { + return -total; + } + + if (model->vocab.tokenizer_clean_spaces) { + text -= total; // restart text + + // first pass: characters ?!., //TODO: where do these characters come from? + const int32_t total1 = total; + total = total ? 1 : 0; + for (int32_t i = 1; i < total1; ++i) { + const char x = text[i]; + if (text[i - 1] == ' ') { + if (x == '?' || x == '!' || x == '.' || x == ',') { // " ?", " !", " .", " ," + total--; // remove space + } + } + text[total++] = x; + } + + // second pass: strip single apostrophe between spaces + const int32_t total2 = total; + total = total ? 1 : 0; + for (int32_t i = 1; i < total2; ++i) { + const char x = text[i]; + if (x == '\'' && i + 1 < total2 && text[i - 1] == ' ' && text[i + 1] == ' ') { // " ' " + total--; // remove prev space + text[++i] = '\0'; // remove next space + } + text[total++] = x; + } + + // third pass: apostrophe contractions //NOTE: this makes sense? + const int32_t total3 = total; + total = total ? 1 : 0; + for (int32_t i = 1; i < total3; ++i) { + const char x = text[i]; + if (text[i - 1] == ' ') { + if (x == '\'' && i + 1 < total3) { + const char x1 = text[i + 1]; + if (x1 == 't' || x1 == 'd') { // " 't", " 'd" + //total--; // remove space + } else if (x1 == 's' || x1 == 'm') { // " 's", " 'm" + total--; // remove space + } else if (i + 2 < total3) { + const char x2 = text[i + 2]; + if ((x1 == 'l' && x2 == 'l')) { // " 'll" + //total--; // remove space + } else if ((x1 == 'r' && x2 == 'e') || (x1 == 'v' && x2 == 'e')) { // " 're", " 've" + total--; // remove space + } else { + //total--; // remove space + } + } else { + //total--; // remove space + } + } + } + text[total++] = x; + } + } + + return total <= text_len_max ? total : -total; +} + // trim whitespace from the beginning and end of a string static std::string trim(const std::string & str) { size_t start = 0; diff --git a/src/unicode.cpp b/src/unicode.cpp index 8692924b9..51daa15af 100644 --- a/src/unicode.cpp +++ b/src/unicode.cpp @@ -232,8 +232,7 @@ static std::vector unicode_regex_split_custom_gpt2(const std::string & t }; auto _get_flags = [&] (const size_t pos) -> codepoint_flags { - static const codepoint_flags undef(codepoint_flags::UNDEFINED); - return (offset_ini <= pos && pos < offset_end) ? unicode_cpt_flags(cpts[pos]) : undef; + return (offset_ini <= pos && pos < offset_end) ? unicode_cpt_flags(cpts[pos]) : codepoint_flags{}; }; size_t _prev_end = offset_ini; @@ -295,9 +294,9 @@ static std::vector unicode_regex_split_custom_gpt2(const std::string & t continue; } // regex: ?[^\s\p{L}\p{N}]+ - if (!(flags2.is_whitespace || flags2.is_letter || flags2.is_number || flags2.is_undefined)) { + if (!(flags2.is_whitespace | flags2.is_letter | flags2.is_number) && flags2.as_uint()) { pos += (cpt == ' '); - while (!(flags2.is_whitespace || flags2.is_letter || flags2.is_number || flags2.is_undefined)) { + while (!(flags2.is_whitespace | flags2.is_letter | flags2.is_number) && flags2.as_uint()) { flags2 = _get_flags(++pos); } _add_token(pos); @@ -351,8 +350,7 @@ static std::vector unicode_regex_split_custom_llama3(const std::string & }; auto _get_flags = [&] (const size_t pos) -> codepoint_flags { - static const codepoint_flags undef(codepoint_flags::UNDEFINED); - return (offset_ini <= pos && pos < offset_end) ? unicode_cpt_flags(cpts[pos]) : undef; + return (offset_ini <= pos && pos < offset_end) ? unicode_cpt_flags(cpts[pos]) : codepoint_flags{}; }; size_t _prev_end = offset_ini; @@ -394,8 +392,8 @@ static std::vector unicode_regex_split_custom_llama3(const std::string & } } - // regex: [^\r\n\p{L}\p{N}]?\p{L}+ //####FIXME: the first \p{L} is correct? - if (!(cpt == '\r' || cpt == '\n' || /*flags.is_letter |*/ flags.is_number)) { + // regex: [^\r\n\p{L}\p{N}]?\p{L}+ + if (!(cpt == '\r' || cpt == '\n' || flags.is_number)) { if (flags.is_letter || _get_flags(pos+1).is_letter) { // one or more letters pos++; while (_get_flags(pos).is_letter) { @@ -421,9 +419,9 @@ static std::vector unicode_regex_split_custom_llama3(const std::string & // regex: ?[^\s\p{L}\p{N}]+[\r\n]* auto flags2 = (cpt == ' ' ? _get_flags(pos+1) : flags); - if (!(flags2.is_whitespace || flags2.is_letter || flags2.is_number || flags2.is_undefined)) { + if (!(flags2.is_whitespace | flags2.is_letter | flags2.is_number) && flags.as_uint()) { pos += (cpt == ' '); - while (!(flags2.is_whitespace || flags2.is_letter || flags2.is_number || flags2.is_undefined)) { + while (!(flags2.is_whitespace | flags2.is_letter | flags2.is_number) && flags2.as_uint()) { flags2 = _get_flags(++pos); } uint32_t cpt2 = _get_cpt(pos); diff --git a/tests/test-tokenizer-0.cpp b/tests/test-tokenizer-0.cpp index d478f1041..1f04b6f34 100644 --- a/tests/test-tokenizer-0.cpp +++ b/tests/test-tokenizer-0.cpp @@ -195,11 +195,11 @@ int main(int argc, char **argv) { const bool add_special = false; for (const auto & test_kv : k_tests) { - const std::vector res = llama_tokenize(ctx, test_kv.first, add_special); + const std::vector res = llama_tokenize(ctx, test_kv.first, add_special, true); printf("\n"); printf("src: '%s'\n", test_kv.first.c_str()); - printf("res: '%s'\n", llama_detokenize_bpe(ctx, res).c_str()); + printf("res: '%s'\n", llama_detokenize(ctx, res).c_str()); printf("tok: "); for (const auto & tok : res) { printf("%d ", tok); @@ -216,8 +216,8 @@ int main(int argc, char **argv) { if (!correct) { fprintf(stderr, "%s : failed test: '%s'\n", __func__, test_kv.first.c_str()); fprintf(stderr, "%s : detokenized to: '%s' instead of '%s'\n", __func__, - llama_detokenize_bpe(ctx, res).c_str(), - llama_detokenize_bpe(ctx, test_kv.second).c_str()); + llama_detokenize(ctx, res).c_str(), + llama_detokenize(ctx, test_kv.second).c_str()); fprintf(stderr, "%s : expected tokens: ", __func__); for (const auto & t : test_kv.second) { fprintf(stderr, "%6d '%s', ", t, llama_token_to_piece(ctx, t).c_str()); @@ -253,7 +253,7 @@ int main(int argc, char **argv) { { const auto t_start = ggml_time_us(); - res = llama_tokenize(ctx, text, add_special); + res = llama_tokenize(ctx, text, add_special, true); const auto t_end = ggml_time_us(); @@ -272,7 +272,7 @@ int main(int argc, char **argv) { } for (const auto & tok : res) { - //ofs << tok << " '" << string_strip(llama_detokenize_bpe(ctx, std::vector{tok})) << "'" << std::endl; + //ofs << tok << " '" << string_strip(llama_detokenize(ctx, std::vector{tok})) << "'" << std::endl; ofs << tok << "\n"; } } diff --git a/tests/test-tokenizer-1-bpe.cpp b/tests/test-tokenizer-1-bpe.cpp index 209a04ad6..9498387e0 100644 --- a/tests/test-tokenizer-1-bpe.cpp +++ b/tests/test-tokenizer-1-bpe.cpp @@ -11,6 +11,7 @@ #include #include #include +#include int main(int argc, char **argv) { if (argc < 2 || argc > 3) { @@ -63,7 +64,10 @@ int main(int argc, char **argv) { } } - GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_BPE); + //GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_BPE); + if (llama_vocab_type(model) != LLAMA_VOCAB_TYPE_BPE) { + return 99; + } #ifdef _WIN32 // We need this for unicode console support @@ -74,7 +78,7 @@ int main(int argc, char **argv) { const int n_vocab = llama_n_vocab(model); for (int i = 0; i < n_vocab; ++i) { - std::string str = llama_detokenize_bpe(ctx, std::vector(1, i)); + std::string str = llama_detokenize(ctx, std::vector(1, i)); try { auto cps = unicode_cpts_from_utf8(str); std::vector tokens = llama_tokenize(ctx, str, false, true); @@ -90,7 +94,7 @@ int main(int argc, char **argv) { fprintf(stderr, "]\n"); return 2; } - std::string check = llama_detokenize_bpe(ctx, tokens); + std::string check = llama_detokenize(ctx, tokens); if (check != str) { fprintf(stderr, "%s : error: token %d detokenizes to '%s'(%zu) but tokenization of this detokenizes to '%s'(%zu)\n", __func__, i, str.c_str(), str.length(), check.c_str(), check.length()); @@ -108,26 +112,23 @@ int main(int argc, char **argv) { std::vector threads(nthread); + std::atomic_int errcode = {}; + for (int i = 0; i < nthread; ++i) { - threads[i] = std::thread([i, nthread, ctx]() { - for (uint32_t cp = i; cp < 0x0010ffff; cp += nthread) { - if (!( // NOLINT - (cp < 0x03 || cp > 0x05) && cp != 0x0b && cp != 0x11 && - (cp < 0x13 || cp > 0x17) && cp != 0x19 && - (cp < 0x1c || cp > 0x1e) && - (cp < 0xd800 || cp > 0xdfff) && - (cp < 0x00040000 || cp >= 0x000e0000) - )) { + threads[i] = std::thread([i, nthread, ctx, &errcode]() { + for (uint32_t cp = i; !errcode && cp < 0x00110000; cp += nthread) { + if ((0x0000D800 <= cp && cp <= 0x0000DFFF) || // surrogates \p{Cs} + (0x00040000 <= cp && cp <= 0x000E0000)) { // undefined \p{Cn} continue; } std::string str = unicode_cpt_to_utf8(cp); std::vector tokens = llama_tokenize(ctx, str, false); - std::string check = llama_detokenize_bpe(ctx, tokens); + std::string check = llama_detokenize(ctx, tokens); if (cp != 9601 && str != check) { - fprintf(stderr, "error: codepoint %x detokenizes to '%s'(%zu) instead of '%s'(%zu)\n", + fprintf(stderr, "error: codepoint 0x%x detokenizes to '%s'(%zu) instead of '%s'(%zu)\n", cp, check.c_str(), check.length(), str.c_str(), str.length()); - std::exit(3); + errcode = 3; } } }); @@ -136,6 +137,10 @@ int main(int argc, char **argv) { for (auto & t : threads) { t.join(); } + + if (errcode) { + return errcode; + } } llama_free_model(model); diff --git a/tests/test-tokenizer-1-spm.cpp b/tests/test-tokenizer-1-spm.cpp index ac2333dda..7ca9e2ca6 100644 --- a/tests/test-tokenizer-1-spm.cpp +++ b/tests/test-tokenizer-1-spm.cpp @@ -11,6 +11,7 @@ #include #include #include +#include int main(int argc, char ** argv) { if (argc < 2) { @@ -51,7 +52,10 @@ int main(int argc, char ** argv) { } } - GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_SPM); + //GGML_ASSERT(llama_vocab_type(model) == LLAMA_VOCAB_TYPE_SPM); + if (llama_vocab_type(model) != LLAMA_VOCAB_TYPE_SPM) { + return 99; + } #ifdef _WIN32 // We need this for unicode console support @@ -62,9 +66,9 @@ int main(int argc, char ** argv) { const int n_vocab = llama_n_vocab(model); for (int i = 0; i < n_vocab; ++i) { - std::string str = llama_detokenize_spm(ctx, std::vector(1, i)); - std::vector tokens = llama_tokenize(ctx, str, false); - std::string check = llama_detokenize_spm(ctx, tokens); + std::string str = llama_detokenize(ctx, std::vector(1, i), true); + std::vector tokens = llama_tokenize(ctx, str, false, true); + std::string check = llama_detokenize(ctx, tokens); if (check != str) { fprintf(stderr, "%s : error: token %d detokenizes to '%s'(%zu) but tokenization of this detokenizes to '%s'(%zu)\n", __func__, i, str.c_str(), str.length(), check.c_str(), check.length()); @@ -78,20 +82,23 @@ int main(int argc, char ** argv) { std::vector threads(nthread); + std::atomic_int errcode = {}; + for (int i = 0; i < nthread; ++i) { - threads[i] = std::thread([i, nthread, ctx]() { - for (uint32_t cp = i; cp < 0x0010ffff; cp += nthread) { - if (cp >= 0xd800 && cp <= 0xdfff) { + threads[i] = std::thread([i, nthread, ctx, &errcode]() { + for (uint32_t cp = i; !errcode && cp < 0x00110000; cp += nthread) { + if ((0x0000D800 <= cp && cp <= 0x0000DFFF) || // surrogates \p{Cs} + (0x00040000 <= cp && cp <= 0x000E0000)) { // undefined \p{Cn} continue; } std::string str = unicode_cpt_to_utf8(cp); - std::vector tokens = llama_tokenize(ctx, str, false); - std::string check = llama_detokenize_spm(ctx, tokens); + std::vector tokens = llama_tokenize(ctx, str, false, true); + std::string check = llama_detokenize(ctx, tokens); if (cp != 9601 && str != check) { - fprintf(stderr, "error: codepoint %x detokenizes to '%s'(%zu) instead of '%s'(%zu)\n", + fprintf(stderr, "error: codepoint 0x%x detokenizes to '%s'(%zu) instead of '%s'(%zu)\n", cp, check.c_str(), check.length(), str.c_str(), str.length()); - std::exit(3); + errcode = 3; } } }); @@ -100,6 +107,10 @@ int main(int argc, char ** argv) { for (auto & t : threads) { t.join(); } + + if(errcode) { + return errcode; + } } llama_free_model(model); diff --git a/tests/test-tokenizer-random.py b/tests/test-tokenizer-random.py index a07c52fb3..48cab8a1e 100644 --- a/tests/test-tokenizer-random.py +++ b/tests/test-tokenizer-random.py @@ -13,7 +13,7 @@ import subprocess import random import unicodedata -from typing import Callable, Iterator +from typing import Iterator import cffi from transformers import AutoTokenizer @@ -24,17 +24,20 @@ logger = logging.getLogger("test-tokenizer-random") class LibLlama: - DEFAULT_PATH_LLAMA_H = "./llama.h" - DEFAULT_PATH_LIBLLAMA = "./build/libllama.so" # CMakeLists.txt: BUILD_SHARED_LIBS ON + DEFAULT_PATH_LLAMA_H = "./include/llama.h" + DEFAULT_PATH_INCLUDES = ["./ggml/include/", "./include/"] + DEFAULT_PATH_LIBLLAMA = "./build/src/libllama.so" # CMakeLists.txt: BUILD_SHARED_LIBS ON - def __init__(self, path_llama_h: str = None, path_libllama: str = None): + def __init__(self, path_llama_h: str = None, path_includes: list[str] = [], path_libllama: str = None): path_llama_h = path_llama_h or self.DEFAULT_PATH_LLAMA_H + path_includes = path_includes or self.DEFAULT_PATH_INCLUDES path_libllama = path_libllama or self.DEFAULT_PATH_LIBLLAMA - (self.ffi, self.lib) = self._load_libllama_cffi(path_llama_h, path_libllama) + (self.ffi, self.lib) = self._load_libllama_cffi(path_llama_h, path_includes, path_libllama) self.lib.llama_backend_init() - def _load_libllama_cffi(self, path_llama_h: str, path_libllama: str): - cmd = ["gcc", "-E", "-P", "-D__restrict=", "-D__attribute__(x)=", "-D__asm__(x)=", path_llama_h] + def _load_libllama_cffi(self, path_llama_h: str, path_includes: list[str], path_libllama: str): + cmd = ["gcc", "-E", "-P", "-D__restrict=", "-D__attribute__(x)=", "-D__asm__(x)="] + cmd += ["-I" + path for path in path_includes] + [path_llama_h] res = subprocess.run(cmd, stdout=subprocess.PIPE) assert (res.returncode == 0) source = res.stdout.decode() @@ -79,6 +82,7 @@ class LibLlamaModel: raise RuntimeError("error: failed to create context for model '%s'" % path_model) n_tokens_max = self.lib.llama_n_ctx(self.ctx) self.token_ids = self.ffi.new("llama_token[]", n_tokens_max) + self.text_buff = self.ffi.new("uint8_t[]", 1024) def free(self): if self.ctx: @@ -89,14 +93,78 @@ class LibLlamaModel: self.model = None self.lib = None - def tokenize(self, text: str, n_tokens_max: int = 0, add_special: bool = False, parse_special: bool = False) -> list[int]: - n_tokens_max = n_tokens_max if n_tokens_max > 0 else len(self.token_ids) + def tokenize(self, text: str, add_special: bool = False, parse_special: bool = False) -> list[int]: text = text.encode("utf-8") - num = self.lib.llama_tokenize(self.model, text, len(text), self.token_ids, n_tokens_max, add_special, parse_special) - if num < 0: - return [] + num = self.lib.llama_tokenize(self.model, text, len(text), self.token_ids, len(self.token_ids), add_special, parse_special) + while num < 0 and len(self.token_ids) < (16 << 20): + self.token_ids = self.ffi.new("llama_token[]", -2 * num) + num = self.lib.llama_tokenize(self.model, text, len(text), self.token_ids, len(self.token_ids), add_special, parse_special) return list(self.token_ids[0:num]) + def detokenize(self, ids: list[int], remove_special: bool = False, unparse_special: bool = False) -> str: + if len(self.token_ids) < len(ids): + self.token_ids = self.ffi.new("llama_token[]", 2 * len(ids)) + for i, id in enumerate(ids): + self.token_ids[i] = id + num = self.lib.llama_detokenize(self.model, self.token_ids, len(ids), self.text_buff, len(self.text_buff), remove_special, unparse_special) + while num < 0 and len(self.text_buff) < (16 << 20): + self.text_buff = self.ffi.new("uint8_t[]", -2 * num) + num = self.lib.llama_detokenize(self.model, self.token_ids, len(ids), self.text_buff, len(self.text_buff), remove_special, unparse_special) + return str(self.ffi.buffer(self.text_buff, num), encoding="utf-8", errors="replace") # replace errors with '\uFFFD' + + +class Tokenizer: + + def encode(self, text: str) -> list[int]: + raise NotImplementedError + + def decode(self, ids: list[int]) -> str: + raise NotImplementedError + + +class TokenizerGroundtruth (Tokenizer): + + def __init__(self, dir_tokenizer: str): + self.model = AutoTokenizer.from_pretrained(dir_tokenizer) + # guess BOS and EOS + ids = self.encode("a") + assert 1 <= len(ids) <= 3 + add_bos_token = len(ids) > 1 and self.model.bos_token_id == ids[0] + add_eos_token = len(ids) > 1 and self.model.eos_token_id == ids[-1] + self.add_bos_token = getattr(self.model, "add_bos_token", add_bos_token) + self.add_eos_token = getattr(self.model, "add_eos_token", add_eos_token) + # build vocab + tokens = list(self.model.get_vocab().values()) + self.vocab = self.model.batch_decode(tokens, skip_special_tokens=True) + self.vocab = list(sorted(self.vocab)) + # tokens and lists + self.special_tokens = list(self.model.all_special_tokens) + self.added_tokens = list(self.model.added_tokens_encoder) + self.bos_token = self.model.bos_token + self.eos_token = self.model.eos_token + + def encode(self, text: str) -> list[int]: + return self.model.encode(text, add_special_tokens=True) + + def decode(self, ids: list[int]) -> str: + return self.model.decode(ids, skip_special_tokens=False) + + +class TokenizerLlamaCpp (Tokenizer): + + libllama: LibLlama = None + + def __init__(self, vocab_file: str): + if not self.libllama: + self.libllama = LibLlama() + self.model = LibLlamaModel(self.libllama, vocab_file, mparams=dict(vocab_only=True), cparams=dict(n_ctx=4096)) + + def encode(self, text: str) -> list[int]: + return self.model.tokenize(text, add_special=True, parse_special=True) + + def decode(self, ids: list[int]) -> str: + return self.model.detokenize(ids, remove_special=False, unparse_special=True) + def generator_custom_text() -> Iterator[str]: """General tests""" @@ -165,19 +233,48 @@ def generator_custom_text_edge_cases() -> Iterator[str]: 'a b', # rstrip phi-3 'a b', # lstrip jina-v2 '\xa0aC', # deepseek + '\u2029 \uA3E4', # deepseek-llm + "a ?", + 'å', # mpt + '\U000ac517', # utf-8 encode error, falcon + '\U000522f4', # utf-8 encode error, starcoder + "abcd", + " abcd", ] -def generator_vocab_words(vocab: list[str]) -> Iterator[str]: +def generator_vocab_words(tokenizer: TokenizerGroundtruth) -> Iterator[str]: """Brute force check all vocab words""" - yield from vocab + yield from tokenizer.vocab -def generator_added_lr_strip(tokenizer) -> Iterator[str]: - WHITESPACES = ["", " ", " ", " "] - special_tokens = list(tokenizer.all_special_tokens) - added_tokens = list(tokenizer.added_tokens_encoder) - all_tokens = list(sorted(set(special_tokens + added_tokens))) +def generator_ascii_lr_strip() -> Iterator[str]: + WHITESPACES = ["", " ", " "] + CHARACTERS = list(chr(i) for i in range(1, 0x80)) + [""] + for char1 in CHARACTERS: + for char2 in CHARACTERS: + for lstrip in WHITESPACES: + for rstrip in WHITESPACES: + yield lstrip + char1 + char2 + rstrip + yield lstrip + char1 + rstrip + char2 + yield char1 + lstrip + char2 + rstrip + + +def generator_apostrophe() -> Iterator[str]: + WHITESPACES = ["", " ", " "] + CHARACTERS = list(chr(i) for i in range(1, 0x80)) + [""] + for char1 in CHARACTERS: + for char2 in CHARACTERS: + for lstrip in WHITESPACES: + for rstrip in WHITESPACES: + yield char1 + lstrip + "'" + rstrip + char2 + yield char1 + char2 + lstrip + "'" + rstrip + "z" + yield "a" + lstrip + "'" + rstrip + char1 + char2 + + +def generator_added_lr_strip(tokenizer: TokenizerGroundtruth) -> Iterator[str]: + WHITESPACES = ["", " ", " ", "\n", "\r\n", "\n\n", "\t", "\t\t"] + all_tokens = list(sorted(set(tokenizer.special_tokens + tokenizer.added_tokens))) for token in all_tokens: for lstrip in WHITESPACES: for rstrip in WHITESPACES: @@ -187,11 +284,9 @@ def generator_added_lr_strip(tokenizer) -> Iterator[str]: yield "a" + lstrip + token + rstrip + "z" -def generator_random_added_tokens(tokenizer, iterations=100) -> Iterator[str]: - special_tokens = list(tokenizer.all_special_tokens) - added_tokens = list(tokenizer.added_tokens_encoder) - separations = [" ", "\n", "\t", "-", "!", "one", "1", "", ""] - all_tokens = list(sorted(set(special_tokens + added_tokens + separations))) +def generator_random_added_tokens(tokenizer: TokenizerGroundtruth, iterations=100) -> Iterator[str]: + separations = [" ", "\n", "\t", "-", "!", "one", "1", "", ""] + all_tokens = list(sorted(set(tokenizer.special_tokens + tokenizer.added_tokens + separations))) rand = random.Random() for m in range(iterations): rand.seed(m) @@ -242,13 +337,13 @@ def generator_unicodes() -> Iterator[str]: def _valid(cpt): if cpt >= 0x30000: # unassigned and supplement­ary return False - if 0x00D800 <= cpt <= 0x00F8FF: # Surrogates - return False - if unicodedata.category(chr(cpt)) == "Cn": + # if cpt == 0x2029: # deepseek-llm + # return False + if unicodedata.category(chr(cpt)) in ("Cn", "Cs", "Co"): # undefined, surrogates, private return False return True - characters = [chr(cpt) for cpt in range(1, MAX_CODEPOINTS) if _valid(cpt)] + characters = [chr(cpt) for cpt in range(0, MAX_CODEPOINTS) if _valid(cpt)] yield from characters @@ -273,11 +368,11 @@ def generator_random_unicodes(iterations=100) -> Iterator[str]: yield "".join(text) -def generator_random_vocab_chars(vocab: list[str], iterations=100) -> Iterator[str]: +def generator_random_vocab_chars(tokenizer: TokenizerGroundtruth, iterations=100) -> Iterator[str]: """Brute force random text with vocab characters""" vocab_chars = set() - for word in vocab: + for word in tokenizer.vocab: vocab_chars.update(word) vocab_chars = list(sorted(vocab_chars)) @@ -288,10 +383,10 @@ def generator_random_vocab_chars(vocab: list[str], iterations=100) -> Iterator[s yield "".join(text) -def generator_random_vocab_words(vocab: list[str], iterations=100) -> Iterator[str]: +def generator_random_vocab_words(tokenizer: TokenizerGroundtruth, iterations=100) -> Iterator[str]: """Brute force random text from vocab words""" - vocab = [w.strip() for w in vocab] + vocab = [w.strip() for w in tokenizer.vocab] yield from vocab rand = random.Random() @@ -307,7 +402,7 @@ def generator_random_vocab_words(vocab: list[str], iterations=100) -> Iterator[s yield "".join(text) -def compare_tokenizers(func_tokenize1: Callable, func_tokenize2: Callable, generator: Iterator[str]): +def compare_tokenizers(tokenizer1: TokenizerGroundtruth, tokenizer2: TokenizerLlamaCpp, generator: Iterator[str]): def find_first_mismatch(ids1: list[int], ids2: list[int]): for i, (a, b) in enumerate(zip(ids1, ids2)): @@ -317,34 +412,67 @@ def compare_tokenizers(func_tokenize1: Callable, func_tokenize2: Callable, gener return -1 return min(len(ids1), len(ids2)) - t_tokenizer1 = 0 - t_tokenizer2 = 0 + def check_detokenizer(text: str, text1: str, text2: str) -> bool: + if text1 == text2: # equal to TokenizerGroundtruth? + return True + # equal to source text? + if tokenizer1.add_bos_token: # remove BOS + if text2.startswith(tokenizer1.bos_token): + text2 = text2[len(tokenizer1.bos_token):] + if tokenizer1.add_eos_token: # remove EOS + if text2.endswith(tokenizer1.eos_token): + text2 = text2[:-len(tokenizer1.eos_token)] + return text == text2 + + t_encode1 = 0 + t_encode2 = 0 + t_decode1 = 0 + t_decode2 = 0 t_start = time.perf_counter() - num_errors = 10 + encode_errors = 0 + decode_errors = 0 + MAX_ERRORS = 10 logger.info("%s: %s" % (generator.__name__, "ini")) for text in generator: + # print(repr(text), text.encode()) # print(repr(text), hex(ord(text[0])), text.encode()) t0 = time.perf_counter() - ids1 = func_tokenize1(text) + ids1 = tokenizer1.encode(text) t1 = time.perf_counter() - ids2 = func_tokenize2(text) + ids2 = tokenizer2.encode(text) t2 = time.perf_counter() - t_tokenizer1 += t1 - t0 - t_tokenizer2 += t2 - t1 - if ids1 != ids2: + text1 = tokenizer1.decode(ids1) + t3 = time.perf_counter() + text2 = tokenizer2.decode(ids1) + t4 = time.perf_counter() + t_encode1 += t1 - t0 + t_encode2 += t2 - t1 + t_decode1 += t3 - t2 + t_decode2 += t4 - t3 + if encode_errors < MAX_ERRORS and ids1 != ids2: i = find_first_mismatch(ids1, ids2) ids1 = list(ids1)[max(0, i - 2) : i + 5 + 1] ids2 = list(ids2)[max(0, i - 2) : i + 5 + 1] - logger.error(" TokenIDs: " + str(ids1)) - logger.error(" Expected: " + str(ids2)) + logger.error(" Expected: " + str(ids1)) + logger.error(" Result: " + str(ids2)) + encode_errors += 1 + logger.error(f" {encode_errors=}") + if decode_errors < MAX_ERRORS and not check_detokenizer(text, text1, text2): + i = find_first_mismatch(text1, text2) + text1 = list(text1[max(0, i - 2) : i + 5 + 1]) + text2 = list(text2[max(0, i - 2) : i + 5 + 1]) + logger.error(" Expected: " + " ".join(hex(ord(x)) for x in text1)) + logger.error(" Result: " + " ".join(hex(ord(x)) for x in text2)) + decode_errors += 1 + logger.error(f" {decode_errors=}") + if encode_errors >= MAX_ERRORS and decode_errors >= MAX_ERRORS: + logger.error(f" EXIT: {encode_errors=} {decode_errors=}") # raise Exception() - num_errors += 1 - if num_errors > 10: - break + break t_total = time.perf_counter() - t_start - logger.info("%s: end, tok1: %.3f tok2: %.3f total: %.3f" % (generator.__name__, t_tokenizer1, t_tokenizer2, t_total)) + logger.info(f"{generator.__name__}: end, {t_encode1=:.3f} {t_encode2=:.3f} {t_decode1=:.3f} {t_decode2=:.3f} {t_total=:.3f}") def main(argv: list[str] = None): @@ -357,74 +485,76 @@ def main(argv: list[str] = None): logging.basicConfig(level = logging.DEBUG if args.verbose else logging.INFO) logger.info(f"VOCABFILE: '{args.vocab_file}'") - model = LibLlamaModel(LibLlama(), args.vocab_file, mparams=dict(vocab_only=True), cparams=dict(n_ctx=4096)) - tokenizer = AutoTokenizer.from_pretrained(args.dir_tokenizer) + tokenizer1 = TokenizerGroundtruth(args.dir_tokenizer) + tokenizer2 = TokenizerLlamaCpp(args.vocab_file) - def func_tokenize1(text: str): - return model.tokenize(text, add_special=True, parse_special=True) + # compare_tokenizers(tokenizer1, tokenizer2, generator_custom_text()) + # compare_tokenizers(tokenizer1, tokenizer2, generator_custom_text_edge_cases()) + compare_tokenizers(tokenizer1, tokenizer2, generator_ascii_lr_strip()) + compare_tokenizers(tokenizer1, tokenizer2, generator_apostrophe()) + compare_tokenizers(tokenizer1, tokenizer2, generator_unicodes()) + compare_tokenizers(tokenizer1, tokenizer2, generator_vocab_words(tokenizer1)) + compare_tokenizers(tokenizer1, tokenizer2, generator_added_lr_strip(tokenizer1)) + # compare_tokenizers(tokenizer1, tokenizer2, generator_random_added_tokens(tokenizer1, 10_000)) + # compare_tokenizers(tokenizer1, tokenizer2, generator_random_chars(10_000)) + # compare_tokenizers(tokenizer1, tokenizer2, generator_random_unicodes(10_000)) + # compare_tokenizers(tokenizer1, tokenizer2, generator_random_vocab_chars(tokenizer1, 10_000)) + # compare_tokenizers(tokenizer1, tokenizer2, generator_random_vocab_words(tokenizer1, 5_000)) - def func_tokenize2(text: str): - return tokenizer.encode(text, add_special_tokens=True) - - ids = func_tokenize2("a") - assert 1 <= len(ids) <= 3 - add_bos_token = len(ids) > 1 and tokenizer.bos_token_id == ids[0] - add_eos_token = len(ids) > 1 and tokenizer.eos_token_id == ids[-1] - tokenizer.add_bos_token = getattr(tokenizer, "add_bos_token", add_bos_token) - tokenizer.add_eos_token = getattr(tokenizer, "add_eos_token", add_eos_token) - - vocab = list(sorted(tokenizer.batch_decode(list(tokenizer.get_vocab().values()), skip_special_tokens=True))) - - compare_tokenizers(func_tokenize1, func_tokenize2, generator_custom_text()) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_custom_text_edge_cases()) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_unicodes()) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_vocab_words(vocab)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_added_lr_strip(tokenizer)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_random_added_tokens(tokenizer, 10_000)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_random_chars(10_000)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_random_unicodes(10_000)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_random_vocab_chars(vocab, 10_000)) - compare_tokenizers(func_tokenize1, func_tokenize2, generator_random_vocab_words(vocab, 5_000)) - - model.free() + tokenizer2.model.free() if __name__ == "__main__": # main() + if True: + logging.basicConfig( + level = logging.DEBUG, + format = "%(asctime)s.%(msecs)03d %(name)s %(levelname)s %(message)s", + datefmt = "%Y-%m-%d %H:%M:%S", + filename = logger.name + ".log", + filemode = "a" + ) logging.basicConfig( level = logging.DEBUG, - format = "%(asctime)s.%(msecs)03d %(name)s %(levelname)s %(message)s", - datefmt = "%Y-%m-%d %H:%M:%S", - filename = logger.name + ".log", - filemode = "a" + format = "%(levelname)s %(message)s", ) path_tokenizers = "./models/tokenizers/" path_vocab_format = "./models/ggml-vocab-%s.gguf" - # import os - # tokenizers = os.listdir(path_tokenizers) tokenizers = [ - # "llama-spm", # SPM - # "phi-3", # SPM - # "bert-bge", # WPM - # "jina-v2-en", # WPM - "gpt-2", # BPE + "llama-spm", # SPM + "phi-3", # SPM + "gemma", # SPM + "gemma-2", # SPM + "baichuan", # SPM + "bert-bge", # WPM + "jina-v2-en", # WPM "llama-bpe", # BPE + "phi-2", # BPE + "deepseek-llm", # BPE + "deepseek-coder", # BPE "falcon", # BPE + "mpt", # BPE "starcoder", # BPE + "gpt-2", # BPE + "stablelm2", # BPE + "refact", # BPE + "qwen2", # BPE + "olmo", # BPE "jina-v2-es", # BPE "jina-v2-de", # BPE - "jina-v2-code", # BPE "smaug-bpe", # BPE - "phi-2", # BPE - "deepseek-coder", # BPE - "deepseek-llm", # BPE + "poro-chat", # BPE + "jina-v2-code", # BPE + "viking", # BPE + "jais", # BPE ] + logger.info("=" * 50) for tokenizer in tokenizers: - logger.info("=" * 50) + logger.info("-" * 50) logger.info(f"TOKENIZER: '{tokenizer}'") vocab_file = path_vocab_format % tokenizer dir_tokenizer = path_tokenizers + "/" + tokenizer