This adds support to llama.cpp to load the model.
Currently missing are changes that are required from convert.py to convert the model correctly. It needs some changes to start reading the JSON configuration for HF models instead of deriving the values by guessing.
Co-authored-by: FNsi <125447286+FNsi@users.noreply.github.com>
1. Add a `LLAMA_SUPPORTS_GPU_OFFLOAD` define to `llama.h` (defined when compiled with CLBlast or cuBLAS)
2. Update the argument handling in the common example code to only show the `-ngl`, `--n-gpu-layers` option when GPU offload is possible.
3. Add an entry for the `-ngl`, `--n-gpu-layers` option to the `main` and `server` examples documentation
4. Update `main` and `server` examples documentation to use the new style dash separator argument format
5. Update the `server` example to use dash separators for its arguments and adds `-ngl` to `--help` (only shown when compiled with appropriate support). It will still support `--memory_f32` and `--ctx_size` for compatibility.
6. Add a warning discouraging use of `--memory-f32` for the `main` and `server` examples `--help` text as well as documentation. Rationale: https://github.com/ggerganov/llama.cpp/discussions/1593#discussioncomment-6004356
- If a given set of values coming along doesn't contain top_k for example, but did before, it would have stayed on the old value, I'm pretty sure. This fixes that.
- I don't know if this could be done a bit prettier by just setting llama.params = gpt_params(); since I'm not sure how the default constructor would react since there's not one defined.
Now passing server params to the help printer so they defaults are ouput.
Bad UTF while streaming now returns a replacement character (\uFFFD)
Changed some error language very slightly.
The JSON now returns extra values, only on `stop` for streaming requests.
New JSON Return Values:
- tokens_predicted (added to streaming)
- seed (just pulls it from params, might return -1)
- prompt (Might be useful)
- generated_text (Full generated response for streaming requests)
sever.cpp left out a few generation parameters and also seems built to assume un-editable chatting with no regens or swipes. I added a simple "reload_ctx" flag that can be passed on generation that will cause the prompt to be reloaded.
Added the one that was supposed to be there.
Apologies for the extra commits, I'm copy pasting from my editor to preserve the two-space indent formatting.
Set `LLAMA_BUILD_SERVER` in workflow so the `server` example gets build. This currently only applies to Windows builds because it seems like only Windows binary artifacts are included in releases.
Add `server` example target to `Makefile` (still uses `LLAMA_BUILD_SERVER` define and does not build by default)
Fix issue where `vdot` binary wasn't removed when running `make clean`.
Fix compile warnings in `server` example.
Add `.hpp` files to trigger workflow (the server example has one).
Improvements to loading the session with `--prompt-cache` in the `main` example.
1. Fix an issue where the `--seed` parameter was ignored when loading a cached prompt.
2. When loading a cached prompt, you previously had to specify the saved prompt (or a prefix of it) again. This pull changes that behavior to default to the prompt that was cached if a prompt wasn't specified by the user.
* xor hack
* block y dim
* loop unrolling
* Fixed cmake LLAMA_CUDA_BY option
* Removed hipblas compatibility code
* Define GGML_CUDA_DMMV_BLOCK_Y if not defined
* Fewer iters, more ops per iter
* Renamed DMMV X/Y compilation options
