| .. |
|
instantiations
|
.
|
2025-09-20 16:09:34 +08:00 |
|
__init__.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
benchmark_attn.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
benchmark_flash_attention_fp8.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
benchmark_split_kv.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
combine.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
copy_sm90_bulk_reduce.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
epilogue_bwd.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
epilogue_fwd.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_api.cpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_attn_interface.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_bwd_kernel_sm80.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_bwd_kernel_sm90.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_bwd_launch_template.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_bwd_postprocess_kernel.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_bwd_preprocess_kernel.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_combine.cu
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_combine_kernel.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_combine_launch_template.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_kernel_sm80.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_kernel_sm90.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
flash_fwd_launch_template.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
generate_kernels.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
heuristics.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
mainloop_bwd_sm80.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
mainloop_bwd_sm90_tma_gmma_ws.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
mainloop_fwd_sm80.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
mainloop_fwd_sm90_tma_gmma_ws.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
mask.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
named_barrier.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
pack_gqa.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
padding.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
paged_kv.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
rotary.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
seqlen.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
setup.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
sm90_pipeline_no_cluster.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
softmax.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
static_switch.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
test_attn_kvcache.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
test_flash_attn.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
test_kvcache.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
test_util.py
|
.
|
2025-09-20 16:09:34 +08:00 |
|
tile_scheduler.hpp
|
.
|
2025-09-20 16:09:34 +08:00 |
|
tile_size.h
|
.
|
2025-09-20 16:09:34 +08:00 |
|
utils.h
|
.
|
2025-09-20 16:09:34 +08:00 |
hailin
c8e8c1e9ff