Tri Dao e45a46a5b7 [Rotary] Implement GPT-J style (interleaved) rotary		2023-03-14 14:35:53 -07:00
..
layers	[Rotary] Implement GPT-J style (interleaved) rotary	2023-03-14 14:35:53 -07:00
losses	Tweak CrossEntropyLoss to take process_group in init	2022-12-27 10:47:43 -08:00
models	[OPT] Load fp16 weights on CPU before moving to GPU	2023-01-22 17:01:32 -08:00
modules	[FusedDense] Support relu, rename FusedDenseGeluDense -> FusedMLP	2023-01-17 18:12:27 -08:00
ops	[LayerNorm] Rename x1 -> residual	2023-01-19 13:07:27 -08:00
utils	[OPT] Load fp16 weights on CPU before moving to GPU	2023-01-22 17:01:32 -08:00
__init__.py	Add missing __init__.py	2022-07-03 02:04:55 -04:00
bert_padding.py	remove numpy dependency	2022-10-06 19:17:15 +02:00
flash_attention.py	Change default dropout value in documentation	2023-01-13 10:50:07 +00:00
flash_attn_interface.py	Fix the case when dout is not contiguous	2022-12-13 13:58:17 -08:00
flash_attn_triton_og.py	Implement FlashAttention in Triton	2022-10-30 18:09:11 -07:00
flash_attn_triton.py	[Triton] Avoid einops repeat by using Tensor.expand	2022-12-14 14:48:41 -08:00
flash_blocksparse_attention.py	Rename src -> flash_attn	2022-06-01 18:50:26 -07:00
flash_blocksparse_attn_interface.py	Rename src -> flash_attn	2022-06-01 18:50:26 -07:00
fused_softmax.py	Add Megatron attention implementation for benchmarking	2022-10-23 23:04:16 -07:00