| .. |
|
flash_attn_triton_amd
|
b518517cb8
[AMD] Triton Backend for ROCm (#1203)
|
vor 3 Monaten |
|
layers
|
8c20cfef49
[Rotary] Support qkv block layout from GQA
|
vor 5 Monaten |
|
losses
|
c7f32a8409
[CrossEntropy] Support precomputed LSE
|
vor 5 Monaten |
|
models
|
30e1ef0f79
minify torch.torch.int32 to torch.int32 (#1237)
|
vor 5 Monaten |
|
modules
|
3f1b4d38e7
Fix: check the type of max_seqlen_k instead of checking max_seqlen twice (#1127)
|
vor 7 Monaten |
|
ops
|
02541ac9e8
[CE] Assert logit_scale > 0
|
vor 1 Monat |
|
utils
|
320fb59487
Update citation
|
vor 9 Monaten |
|
__init__.py
|
5231d95fe1
Drop Pytorch 2.1
|
vor 1 Monat |
|
bert_padding.py
|
30e1ef0f79
minify torch.torch.int32 to torch.int32 (#1237)
|
vor 5 Monaten |
|
flash_attn_interface.py
|
d57f826835
Expose `zero_tensors` arg in varlen functions (#1433)
|
vor 1 Monat |
|
flash_attn_triton.py
|
f1a73d0740
Run isort and black on python files
|
vor 1 Jahr |
|
flash_attn_triton_og.py
|
f1a73d0740
Run isort and black on python files
|
vor 1 Jahr |
|
flash_blocksparse_attention.py
|
cdbbe844b1
minor changes to unpad_input test util func
|
vor 5 Monaten |
|
flash_blocksparse_attn_interface.py
|
f1a73d0740
Run isort and black on python files
|
vor 1 Jahr |
|
fused_softmax.py
|
f1a73d0740
Run isort and black on python files
|
vor 1 Jahr |
|
pyproject.toml
|
73bd3f3bbb
Move pyproject.toml to flash-attn and tests dir to avoid PEP 517
|
vor 1 Jahr |
Tri Dao
02541ac9e8
[CE] Assert logit_scale > 0