flash-attention/hopper
ganeshcolfax 1899c970c8
Changes For FP8 (#1075)
* adding files for fp8 changes.

* removed contiguous check.

* enable all tests except odd-seq-lengths, where it crashes now.

* undid clang formatting.

* change to correct tile size for headdim=128.

* fixed odd-seq-len-k.

* minor formatting.

* minor reformatting.

---------

Co-authored-by: Tri Dao <tridao@users.noreply.github.com>
2024-07-23 13:51:14 -07:00
..
__init__.py FA3 initial code release 2024-07-11 09:53:36 -07:00
benchmark_attn.py Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 2024-07-22 21:32:41 -07:00
benchmark_flash_attention_fp8.py Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
benchmark_flash_attention.py Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
block_info.h FA3 initial code release 2024-07-11 09:53:36 -07:00
epilogue_fwd_sm90_tma.hpp Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_api.cpp Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_attn_interface.py Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_bwd_hdim64_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_bwd_hdim128_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_bwd_hdim256_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_bwd_kernel.h FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_bwd_launch_template.h Remove torchlib dependency from cpp files (#1083) 2024-07-22 16:47:09 -07:00
flash_bwd_preprocess_kernel.h FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_fwd_hdim64_bf16_sm90.cu [FA3] BF16 forward 2024-07-14 23:39:46 -07:00
flash_fwd_hdim64_fp8_sm90.cu Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_fwd_hdim64_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_fwd_hdim128_bf16_sm90.cu [FA3] BF16 forward 2024-07-14 23:39:46 -07:00
flash_fwd_hdim128_fp8_sm90.cu Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_fwd_hdim128_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_fwd_hdim256_bf16_sm90.cu [FA3] BF16 forward 2024-07-14 23:39:46 -07:00
flash_fwd_hdim256_fp8_sm90.cu Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash_fwd_hdim256_fp16_sm90.cu FA3 initial code release 2024-07-11 09:53:36 -07:00
flash_fwd_kernel.h Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 2024-07-22 21:32:41 -07:00
flash_fwd_launch_template.h Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
flash.h Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 2024-07-22 21:32:41 -07:00
kernel_traits.h Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
mainloop_fwd_sm90_tma_gmma_ws.hpp Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
named_barrier.hpp [FA3] BF16 forward 2024-07-14 23:39:46 -07:00
seq_len.h Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 2024-07-22 21:32:41 -07:00
setup.py Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
softmax.h FA3 initial code release 2024-07-11 09:53:36 -07:00
static_switch.h Add var-seq-len to FA3 fp16 / bf16 fwd (#1072) 2024-07-22 21:32:41 -07:00
test_flash_attn.py Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00
tile_scheduler.hpp [FA3] BF16 forward 2024-07-14 23:39:46 -07:00
utils.h Changes For FP8 (#1075) 2024-07-23 13:51:14 -07:00