├── .github └── workflows │ ├── scripts │ ├── build.sh │ ├── cuda-install.sh │ ├── cutlass-install.sh │ ├── env.sh │ └── pytorch-install.sh │ └── wheels.yaml ├── .gitignore ├── LICENSE ├── MANIFEST.in ├── README.md ├── assets ├── flute-logo.png └── intro-figure.jpg ├── examples └── learnable_scales_eval.ipynb ├── flute ├── __init__.py ├── codegen_utils.py ├── csrc │ ├── config.hpp │ ├── conversion_utils.hpp │ ├── cutlass_extensions_bf16.h │ ├── hadamard_transform.cpp │ ├── hadamard_transform_cuda.cu │ ├── marlin_utils.hpp │ ├── packbits_utils.hpp │ ├── qgemm.cpp │ ├── qgemm_kernel.hpp │ ├── qgemm_kernel_example.cu │ ├── qgemm_kernel_raw_generated.cu │ └── tile_scheduler_utils.hpp ├── data │ ├── qgemm_kernel_raw_generated_configs.pth │ ├── qgemm_kernel_raw_tuned_configs.no-M.pth │ └── qgemm_kernel_raw_tuned_configs.pth ├── integrations │ ├── __init__.py │ ├── base.py │ ├── bitsandbytes.py │ ├── higgs.py │ ├── huggingface.py │ ├── learnable.py │ ├── vllm.py │ └── vllm_utils.py ├── nf_utils.py ├── ops.py ├── packbits_utils.py ├── tune.py └── utils.py ├── pyproject.toml ├── requirements.txt ├── scripts └── codegen_tuned.sh ├── setup.py └── tests ├── __init__.py ├── higgs.py ├── kernel.py ├── shapes.py ├── vllm.py └── vllm.sh /.github/workflows/scripts/build.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/scripts/build.sh -------------------------------------------------------------------------------- /.github/workflows/scripts/cuda-install.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/scripts/cuda-install.sh -------------------------------------------------------------------------------- /.github/workflows/scripts/cutlass-install.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/scripts/cutlass-install.sh -------------------------------------------------------------------------------- /.github/workflows/scripts/env.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/scripts/env.sh -------------------------------------------------------------------------------- /.github/workflows/scripts/pytorch-install.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/scripts/pytorch-install.sh -------------------------------------------------------------------------------- /.github/workflows/wheels.yaml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.github/workflows/wheels.yaml -------------------------------------------------------------------------------- /.gitignore: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/.gitignore -------------------------------------------------------------------------------- /LICENSE: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/LICENSE -------------------------------------------------------------------------------- /MANIFEST.in: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/MANIFEST.in -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/README.md -------------------------------------------------------------------------------- /assets/flute-logo.png: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/assets/flute-logo.png -------------------------------------------------------------------------------- /assets/intro-figure.jpg: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/assets/intro-figure.jpg -------------------------------------------------------------------------------- /examples/learnable_scales_eval.ipynb: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/examples/learnable_scales_eval.ipynb -------------------------------------------------------------------------------- /flute/__init__.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/__init__.py -------------------------------------------------------------------------------- /flute/codegen_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/codegen_utils.py -------------------------------------------------------------------------------- /flute/csrc/config.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/config.hpp -------------------------------------------------------------------------------- /flute/csrc/conversion_utils.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/conversion_utils.hpp -------------------------------------------------------------------------------- /flute/csrc/cutlass_extensions_bf16.h: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/cutlass_extensions_bf16.h -------------------------------------------------------------------------------- /flute/csrc/hadamard_transform.cpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/hadamard_transform.cpp -------------------------------------------------------------------------------- /flute/csrc/hadamard_transform_cuda.cu: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/hadamard_transform_cuda.cu -------------------------------------------------------------------------------- /flute/csrc/marlin_utils.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/marlin_utils.hpp -------------------------------------------------------------------------------- /flute/csrc/packbits_utils.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/packbits_utils.hpp -------------------------------------------------------------------------------- /flute/csrc/qgemm.cpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/qgemm.cpp -------------------------------------------------------------------------------- /flute/csrc/qgemm_kernel.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/qgemm_kernel.hpp -------------------------------------------------------------------------------- /flute/csrc/qgemm_kernel_example.cu: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/qgemm_kernel_example.cu -------------------------------------------------------------------------------- /flute/csrc/qgemm_kernel_raw_generated.cu: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/qgemm_kernel_raw_generated.cu -------------------------------------------------------------------------------- /flute/csrc/tile_scheduler_utils.hpp: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/csrc/tile_scheduler_utils.hpp -------------------------------------------------------------------------------- /flute/data/qgemm_kernel_raw_generated_configs.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/data/qgemm_kernel_raw_generated_configs.pth -------------------------------------------------------------------------------- /flute/data/qgemm_kernel_raw_tuned_configs.no-M.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/data/qgemm_kernel_raw_tuned_configs.no-M.pth -------------------------------------------------------------------------------- /flute/data/qgemm_kernel_raw_tuned_configs.pth: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/data/qgemm_kernel_raw_tuned_configs.pth -------------------------------------------------------------------------------- /flute/integrations/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /flute/integrations/base.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/base.py -------------------------------------------------------------------------------- /flute/integrations/bitsandbytes.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/bitsandbytes.py -------------------------------------------------------------------------------- /flute/integrations/higgs.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/higgs.py -------------------------------------------------------------------------------- /flute/integrations/huggingface.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/huggingface.py -------------------------------------------------------------------------------- /flute/integrations/learnable.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/learnable.py -------------------------------------------------------------------------------- /flute/integrations/vllm.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/vllm.py -------------------------------------------------------------------------------- /flute/integrations/vllm_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/integrations/vllm_utils.py -------------------------------------------------------------------------------- /flute/nf_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/nf_utils.py -------------------------------------------------------------------------------- /flute/ops.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/ops.py -------------------------------------------------------------------------------- /flute/packbits_utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/packbits_utils.py -------------------------------------------------------------------------------- /flute/tune.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/tune.py -------------------------------------------------------------------------------- /flute/utils.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/flute/utils.py -------------------------------------------------------------------------------- /pyproject.toml: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/pyproject.toml -------------------------------------------------------------------------------- /requirements.txt: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/requirements.txt -------------------------------------------------------------------------------- /scripts/codegen_tuned.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/scripts/codegen_tuned.sh -------------------------------------------------------------------------------- /setup.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/setup.py -------------------------------------------------------------------------------- /tests/__init__.py: -------------------------------------------------------------------------------- 1 | -------------------------------------------------------------------------------- /tests/higgs.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/tests/higgs.py -------------------------------------------------------------------------------- /tests/kernel.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/tests/kernel.py -------------------------------------------------------------------------------- /tests/shapes.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/tests/shapes.py -------------------------------------------------------------------------------- /tests/vllm.py: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/tests/vllm.py -------------------------------------------------------------------------------- /tests/vllm.sh: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/HanGuo97/flute/HEAD/tests/vllm.sh --------------------------------------------------------------------------------