[Integration] Upload tutorial for making a bitnet ckpt for vLLM #135
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request includes several significant changes to the
integration/BitNet
directory, focusing on enhancing the documentation, improving scripts for model checkpoints, and refining quantization utilities. The most important changes include adding new sections to the README, introducing scripts for generating model checkpoints, and updating quantization functions.Documentation Updates:
integration/BitNet/README.md
: Added sections for "Latest News" and "Make Checkpoints for vLLM" with instructions and scripts for generating model checkpoints.Script Enhancements:
integration/BitNet/maint/generate_bitnet_model_bitblas_format.sh
: Added a new script for generating BitNet model checkpoints in BitBLAS format.integration/BitNet/maint/generate_bitnet_model_native_format.sh
: Added a new script for generating BitNet model checkpoints in native format.integration/BitNet/maint/create_bitblas_ckpt.py
: Renamed and updated the script to use argparse for handling input arguments. [1] [2]Code Improvements:
integration/BitNet/utils_quant.py
: Added@torch.compile
decorator to quantization functions and refactored theforward
method to usepost_quant_process
. [1] [2] [3]Miscellaneous:
integration/BitNet/vllm_workspace/inference_with_compress_format.py
: Added a script for inference with compressed format models.integration/BitNet/vllm_workspace/inference_with_native_format.py
: Added a script for inference with native format models.integration/BitNet/vllm_workspace/utils.py
: Added utility functions for comparing model outputs and log probabilities.