build.toml

[general]
name = "paged_attention"

[torch]
src = [
  "torch-ext/torch_binding.cpp",
  "torch-ext/torch_binding.h"
]

[kernel.cuda_utils]
cuda-capabilities = ["7.0", "7.2", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0"]
src = [
  "cuda-utils/cuda_utils_kernels.cu",
]
depends = []
 

[kernel.paged_attention]
cuda-capabilities = ["7.0", "7.2", "7.5", "8.0", "8.6", "8.7", "8.9", "9.0", "10.0", "10.1", "12.0"]
src = [
  "paged-attention/attention/attention_dtypes.h",
  "paged-attention/attention/attention_generic.cuh",
  "paged-attention/attention/attention_kernels.cuh",
  "paged-attention/attention/attention_utils.cuh",
  "paged-attention/attention/dtype_bfloat16.cuh",
  "paged-attention/attention/dtype_float16.cuh",
  "paged-attention/attention/dtype_float32.cuh",
  "paged-attention/attention/dtype_fp8.cuh",
  "paged-attention/attention/paged_attention_v1.cu",
  "paged-attention/attention/paged_attention_v2.cu",
  "paged-attention/cache_kernels.cu",
  "paged-attention/cuda_compat.h",
  "paged-attention/dispatch_utils.h",
  "paged-attention/quantization/fp8/amd/hip_float8.h",
  "paged-attention/quantization/fp8/amd/hip_float8_impl.h",
  "paged-attention/quantization/fp8/amd/quant_utils.cuh",
  "paged-attention/quantization/fp8/nvidia/quant_utils.cuh",
]
include = [ "." ]
depends = [ "torch" ]