vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

merge master

Browse source
This commit is contained in:
toyer 2024-07-07 10:39:39 +00:00
commit 223eb18788
119 changed files with 5083 additions and 1842 deletions
Download patch file Download diff file Expand all files Collapse all files

12
.gitignore vendored
View file

@ -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

6
CMakeLists.txt
View file

@ -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

30
CONTRIBUTING.md
View file

@ -1,14 +1,24 @@
# Contributing Guidelines
# Pull requests
## Checklist
- Always squash-merge the PR before merging
- Use the following format for your final commit: `<module> : <commit title> (#<issue_number>)`. 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: `<module> : <commit title> (#<issue_number>)`. For example: `utils : Fix typo in utils.py (#1234)`

766
README.md
View file

@ -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)
----
<details>
<summary>Table of Contents</summary>
<ol>
<li>
<a href="#description">Description</a>
</li>
<li>
<a href="#usage">Usage</a>
<ul>
<li><a href="#get-the-code">Get the Code</a></li>
<li><a href="#build">Build</a></li>
<li><a href="#blas-build">BLAS Build</a></li>
<li><a href="#prepare-and-quantize">Prepare and Quantize</a></li>
<li><a href="#run-the-quantized-model">Run the quantized model</a></li>
<li><a href="#memorydisk-requirements">Memory/Disk Requirements</a></li>
<li><a href="#quantization">Quantization</a></li>
<li><a href="#interactive-mode">Interactive mode</a></li>
<li><a href="#constrained-output-with-grammars">Constrained output with grammars</a></li>
<li><a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a></li>
<li><a href="#seminal-papers-and-background-on-the-models">Seminal papers and background on the models</a></li>
<li><a href="#perplexity-measuring-model-quality">Perplexity (measuring model quality)</a></li>
<li><a href="#android">Android</a></li>
<li><a href="#docker">Docker</a></li>
</ul>
</li>
<li><a href="#contributing">Contributing</a></li>
<li><a href="#coding-guidelines">Coding guidelines</a></li>
<li><a href="#docs">Docs</a></li>
</ol>
</details>
## 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:
<details>
<summary>Typical run using LLaMA v2 13B on M2 Ultra</summary>
```
$ 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
```
</details>
<details>
<summary>Demo of running both LLaMA-7B and whisper.cpp on a single M1 Pro MacBook</summary>
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
</details>
## 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:
If you want a more ChatGPT-like experience, you can run in conversation mode by passing `-cnv` as a parameter:
```bash
make
llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv
# 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!
```
- 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`:
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
cmake -B build
cmake --build build --config Release
./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --chat-template chatml
```
**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):
You can also use your own template via in-prefix, in-suffix and reverse-prompt parameters:
```bash
cmake -B build -DCMAKE_BUILD_TYPE=Debug
cmake --build build
./llama-cli -m your_model.gguf -p "You are a helpful assistant" -cnv --in-prefix 'User: ' --reverse-prompt 'User:'
```
2. Multi-config generators (`-G` param set to Visual Studio, XCode...):
### Web server
[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.
Example usage:
```bash
cmake -B build -G "Xcode"
cmake --build build --config Debug
./llama-server -m your_model.gguf --port 8080
# Basic web UI can be accessed via browser: http://localhost:8080
# Chat completion endpoint: http://localhost:8080/v1/chat/completions
```
- 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 '<nixpkgs>' --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=<directory-you-just-found>`, so something
like:
```bash
HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \
HIP_DEVICE_LIB_PATH=<directory-you-just-found> \
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 <a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a> 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.
```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
<folder containing weights and tokenizer json> vocab.json
# [Optional] for PyTorch .bin models like Mistral-7B
ls ./models
<folder containing weights and tokenizer json>
# 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.
### 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
```
### 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)
### 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:
```
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 <a href="#obtaining-and-using-the-facebook-llama-2-model">Obtaining and using the Facebook LLaMA 2 model</a> 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=<your_ndk_directory>
$ 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)

8
ci/run.sh
View file

@ -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"

141
common/common.cpp
View file

@ -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<struct llama_model *, struct llama_context *> 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<struct llama_model *, struct llama_context *> llama_init_from_gpt_par
if (params.warmup) {
LOG("warming up the model with an empty run\n");
std::vector<llama_token> tmp = { llama_token_bos(model), llama_token_eos(model), };
std::vector<llama_token> 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, decltype(&curl_easy_cleanup)> 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<std::future<bool>> 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_token> llama_tokenize(
}
std::string llama_token_to_piece(const struct llama_context * ctx, llama_token token, bool special) {
std::vector<char> 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);
}
return std::string(result.data(), result.size());
}
std::string llama_detokenize_spm(llama_context * ctx, const std::vector<llama_token> & 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);
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);
}
result += piece;
return piece;
}
return result;
std::string llama_detokenize(llama_context * ctx, const std::vector<llama_token> & 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
}
std::string llama_detokenize_bpe(llama_context * ctx, const std::vector<llama_token> & 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) {

23
common/common.h
View file

@ -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<struct llama_model *, struct llama_context *> 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<llama_token> & 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<llama_token> & tokens);
const std::vector<llama_token> & tokens,
bool special = true);
// Uses the value from the model metadata if possible, otherwise
// defaults to true when model type is SPM, otherwise false.

231
convert-hf-to-gguf.py → convert_hf_to_gguf.py
View file

@ -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
@ -2865,6 +2963,10 @@ class T5Model(Model):
from sentencepiece import SentencePieceProcessor
from sentencepiece import sentencepiece_model_pb2 as 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():
@ -2872,10 +2974,18 @@ class T5Model(Model):
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"])
@ -2961,11 +3074,16 @@ 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.")
# 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)

46
convert-hf-to-gguf-update.py → convert_hf_to_gguf_update.py
View file

@ -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 <huggingface_token>
# python3 convert_hf_to_gguf_update.py <huggingface_token>
#
# - 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 <huggingface_token>")
logger.info("Usage: python convert_hf_to_gguf_update.py <huggingface_token>")
sys.exit(1)
else:
logger.info("Usage: python convert-hf-to-gguf-update.py <huggingface_token>")
logger.info("Usage: python convert_hf_to_gguf_update.py <huggingface_token>")
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
@ -143,6 +151,9 @@ for model in models:
# create the tokenizer
try:
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}")
@ -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,
]
@ -301,6 +314,9 @@ for model in models:
# create the tokenizer
try:
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}")
@ -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")

0
convert-llama-ggml-to-gguf.py → convert_llama_ggml_to_gguf.py
View file

56
docs/android.md Normal file
View file

@ -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=<your_ndk_directory>
$ 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

0
docs/BLIS.md → docs/backend/BLIS.md
View file

0
README-sycl.md → docs/backend/SYCL.md
View file

288
docs/build.md Normal file
View file

@ -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=<directory-you-just-found>`, so something
like:
```bash
HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -p)" \
HIP_DEVICE_LIB_PATH=<directory-you-just-found> \
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)

4
docs/HOWTO-add-model.md → docs/development/HOWTO-add-model.md
View file

@ -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.

0
docs/debugging-tests.md → docs/development/debugging-tests.md
View file

0
docs/llama-star/idea-arch.key → docs/development/llama-star/idea-arch.key
View file

0
docs/llama-star/idea-arch.pdf → docs/development/llama-star/idea-arch.pdf
View file

0
docs/token_generation_performance_tips.md → docs/development/token_generation_performance_tips.md
View file

86
docs/docker.md Normal file
View file

@ -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
```

39
docs/install.md Normal file
View file

@ -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 '<nixpkgs>' --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.

3
examples/batched.swift/Sources/main.swift
View file

@ -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)

34
examples/batched/batched.cpp
View file

@ -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<llama_seq_id> 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);

0
examples/convert-legacy-llama.py → examples/convert_legacy_llama.py
View file

0
examples/finetune/convert-finetune-checkpoint-to-gguf.py → examples/finetune/convert_finetune_checkpoint_to_gguf.py
View file

2
examples/json-schema-pydantic-example.py → examples/json_schema_pydantic_example.py
View file

@ -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

4
examples/llama.swiftui/llama.cpp.swift/LibLlama.swift
View file

@ -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<Int8>.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 {

14
examples/llava/MobileVLM-README.md
View file

@ -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`

20
examples/llava/README.md
View file

@ -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:

0
examples/llava/convert-image-encoder-to-gguf.py → examples/llava/convert_image_encoder_to_gguf.py
View file

0
examples/llava/llava-surgery.py → examples/llava/llava_surgery.py
View file

0
examples/llava/llava-surgery-v2.py → examples/llava/llava_surgery_v2.py
View file

4
examples/llava/requirements.txt
View file

@ -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

38
examples/main/main.cpp
View file

@ -38,6 +38,7 @@ static std::vector<llama_token> * g_input_tokens;
static std::ostringstream * g_output_ss;
static std::vector<llama_token> * g_output_tokens;
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());
@ -100,6 +101,7 @@ static void sigint_handler(int signo) {
if (signo == SIGINT) {
if (!is_interacting && g_params->interactive) {
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);
}
// 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);
if (!llama_model_has_encoder(model)) {
GGML_ASSERT(llama_add_eos_token(model) != 1);
}
LOG("add_bos: %d\n", add_bos);
std::vector<llama_token> 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());

3
examples/passkey/README.md
View file

@ -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

9
examples/perplexity/perplexity.cpp
View file

@ -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) {

0
examples/pydantic-models-to-grammar-examples.py → examples/pydantic_models_to_grammar_examples.py
View file

89
examples/quantize/README.md
View file

@ -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
<folder containing weights and tokenizer json> vocab.json
# [Optional] for PyTorch .bin models like Mistral-7B
ls ./models
<folder containing weights and tokenizer json>
# 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)
-- | --

0
examples/regex-to-grammar.py → examples/regex_to_grammar.py
View file

6
examples/server/README.md
View file

@ -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:*

13
examples/server/server.cpp
View file

@ -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<char> 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");

0
examples/server-embd.py → examples/server_embd.py
View file

8
examples/tokenize/tokenize.cpp
View file

@ -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);

0
examples/train-text-from-scratch/convert-train-checkpoint-to-gguf.py → examples/train-text-from-scratch/convert_train_checkpoint_to_gguf.py
View file

3
ggml/src/CMakeLists.txt
View file

@ -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()

4
ggml/src/ggml-cuda/common.cuh
View file

@ -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

18
ggml/src/ggml-cuda/fattn-common.cuh
View file

@ -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<T, half>::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);

8
ggml/src/ggml-cuda/mmq.cu
View file

@ -59,6 +59,12 @@ void ggml_cuda_op_mul_mat_q(
case GGML_TYPE_Q6_K:
mul_mat_q_case<GGML_TYPE_Q6_K>(ctx, args, stream);
break;
case GGML_TYPE_IQ4_XS:
mul_mat_q_case<GGML_TYPE_IQ4_XS>(ctx, args, stream);
break;
case GGML_TYPE_IQ4_NL:
mul_mat_q_case<GGML_TYPE_IQ4_NL>(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:

208
ggml/src/ggml-cuda/mmq.cuh
View file

@ -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;
@ -97,6 +101,8 @@ static constexpr __host__ __device__ tile_x_sizes mmq_get_dp4a_tile_x_sizes(ggml
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};
}
@ -133,6 +139,8 @@ static constexpr __host__ __device__ int mmq_get_mma_tile_x_k(ggml_type type) {
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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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 <int mmq_y, int nwarps, bool need_check> 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<int mmq_x, int mmq_y, int nwarps, bool need_check>
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<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_Q6_K> {
static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q6_K_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
};
template <int mmq_x, int mmq_y, int nwarps, bool need_check>
struct mmq_type_traits<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_IQ4_NL> {
static constexpr int vdr = VDR_IQ4_NL_Q8_1_MMQ;
static constexpr load_tiles_mmq_t load_tiles = load_tiles_iq4_nl<mmq_y, nwarps, need_check>;
static constexpr vec_dot_mmq_t vec_dot_mma = vec_dot_q5_0_q8_1_mma<mmq_x, mmq_y, nwarps>;
static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
};
template <int mmq_x, int mmq_y, int nwarps, bool need_check>
struct mmq_type_traits<mmq_x, mmq_y, nwarps, need_check, GGML_TYPE_IQ4_XS> {
static constexpr int vdr = VDR_IQ4_XS_Q8_1_MMQ;
static constexpr load_tiles_mmq_t load_tiles = load_tiles_iq4_xs<mmq_y, nwarps, need_check>;
static constexpr vec_dot_mmq_t vec_dot_mma = vec_dot_q5_0_q8_1_mma<mmq_x, mmq_y, nwarps>;
static constexpr vec_dot_mmq_t vec_dot_dp4a = vec_dot_q5_0_q8_1_dp4a<mmq_x, mmq_y, nwarps>;
};
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);
// -------------------------------------------------------------------------------------------------------------------------

3
ggml/src/ggml-cuda/template-instances/generate_cu_files.py
View file

@ -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.

5
ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_nl.cu Normal file
View file

@ -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);

5
ggml/src/ggml-cuda/template-instances/mmq-instance-iq4_xs.cu Normal file
View file

@ -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);

43
ggml/src/ggml-cuda/vecdotq.cuh
View file

@ -1,36 +1,8 @@
#include "common.cuh"
#include <cstdint>
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) {

275
ggml/src/ggml-sycl.cpp
View file

@ -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 <bool vals_smem, int ncols_template, int block_size_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 <bool vals_smem, int ncols_template, int block_size_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<float, 1> 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<vals_smem, ncols_template, block_size_template>(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<sycl::info::device::local_mem_size>();
if (n_local_scratch*sizeof(float) < local_mem_size) {
if (ncols_x > max_block_size) {
soft_max_f32_submitter<true, 0, 0>(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<true, 32, 32>(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<true, 64, 64>(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<true, 128, 128>(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<true, 256, 256>(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<true, 512, 512>(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<true, 1024, 1024>(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<true, 2048, 1024>(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<true, 4096, 1024>(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<true, 0, 0>(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<false, 0, 0>(x, mask, dst, ncols_x, nrows_y, scale,
max_bias, m0, m1, n_head_log2, block_nums,
block_dims, WARP_SIZE, stream);
}
}
template <typename T>
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<sycl::half> 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:

1
ggml/src/ggml-sycl/backend.hpp
View file

@ -21,5 +21,6 @@
#include "mmvq.hpp"
#include "rope.hpp"
#include "norm.hpp"
#include "softmax.hpp"
#endif // GGML_SYCL_BACKEND_HPP

15
ggml/src/ggml-sycl/common.hpp
View file

@ -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<float, GGML_SYCL_MAX_DEVICES> 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,

31
ggml/src/ggml-sycl/dmmv.cpp
View file

@ -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);
});
}

10
ggml/src/ggml-sycl/dpct/helper.hpp
View file

@ -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):

26
ggml/src/ggml-sycl/norm.cpp
View file

@ -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;

1
ggml/src/ggml-sycl/presets.hpp
View file

@ -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

250
ggml/src/ggml-sycl/softmax.cpp Normal file
View file

@ -0,0 +1,250 @@
#include "norm.hpp"
template <bool vals_smem, int ncols_template, int block_size_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 <bool vals_smem, int ncols_template, int block_size_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<float, 1> 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<vals_smem, ncols_template, block_size_template>(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<sycl::info::device::local_mem_size>();
if (n_local_scratch*sizeof(float) < local_mem_size) {
if (ncols_x > max_block_size) {
soft_max_f32_submitter<true, 0, 0>(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<true, 32, 32>(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<true, 64, 64>(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<true, 128, 128>(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<true, 256, 256>(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<true, 512, 512>(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<true, 1024, 1024>(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<true, 2048, 1024>(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<true, 4096, 1024>(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<true, 0, 0>(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<false, 0, 0>(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);
}

24
ggml/src/ggml-sycl/softmax.hpp Normal file
View file

@ -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

2
ggml/src/ggml.c
View file

@ -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

10
gguf-py/README.md
View file

@ -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:

15
gguf-py/gguf/constants.py
View file

@ -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,

15
gguf-py/gguf/gguf_writer.py
View file

@ -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:
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:
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:
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)

14
gguf-py/gguf/tensor_mapping.py
View file

@ -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: (

17
gguf-py/scripts/__init__.py
View file

@ -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

0
gguf-py/scripts/gguf-convert-endian.py → gguf-py/scripts/gguf_convert_endian.py
View file

0
gguf-py/scripts/gguf-dump.py → gguf-py/scripts/gguf_dump.py
View file

0
gguf-py/scripts/gguf-new-metadata.py → gguf-py/scripts/gguf_new_metadata.py
View file

0
gguf-py/scripts/gguf-set-metadata.py → gguf-py/scripts/gguf_set_metadata.py
View file

41
include/llama.h
View file

@ -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

6
models/ggml-vocab-bert-bge.gguf.inp
View file

@ -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__

3
models/ggml-vocab-bert-bge.gguf.out
View file

@ -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

6
models/ggml-vocab-command-r.gguf.inp
View file

@ -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__

3
models/ggml-vocab-command-r.gguf.out
View file

@ -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

6
models/ggml-vocab-deepseek-coder.gguf.inp
View file

@ -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__

3
models/ggml-vocab-deepseek-coder.gguf.out
View file

@ -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

6
models/ggml-vocab-deepseek-llm.gguf.inp
View file

@ -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__

3
models/ggml-vocab-deepseek-llm.gguf.out
View file

@ -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

6
models/ggml-vocab-falcon.gguf.inp
View file

@ -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__

3
models/ggml-vocab-falcon.gguf.out
View file

@ -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

6
models/ggml-vocab-gpt-2.gguf.inp
View file

@ -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__

3
models/ggml-vocab-gpt-2.gguf.out
View file

@ -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

8
models/ggml-vocab-llama-bpe.gguf.inp
View file

@ -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__

4
models/ggml-vocab-llama-bpe.gguf.out
View file

@ -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

6
models/ggml-vocab-llama-spm.gguf.inp
View file

@ -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__

3
models/ggml-vocab-llama-spm.gguf.out
View file

@ -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

6
models/ggml-vocab-mpt.gguf.inp
View file

@ -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__

3
models/ggml-vocab-mpt.gguf.out
View file

@ -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

6
models/ggml-vocab-phi-3.gguf.inp
View file

@ -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__

3
models/ggml-vocab-phi-3.gguf.out
View file

@ -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

6
models/ggml-vocab-qwen2.gguf.inp
View file

@ -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__

3
models/ggml-vocab-qwen2.gguf.out
View file

@ -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

6
models/ggml-vocab-refact.gguf.inp
View file

@ -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__

3
models/ggml-vocab-refact.gguf.out
View file

@ -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

6
models/ggml-vocab-starcoder.gguf.inp
View file

@ -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__

3
models/ggml-vocab-starcoder.gguf.out
View file

@ -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

1197
poetry.lock generated Normal file
View file

File diff suppressed because it is too large Load diff

Some files were not shown because too many files have changed in this diff Show more