Ying Zhang 496fdc4f6c Add seqused_q in fwd / bwd and seqused_k in bwd. hai 4 meses
..
layers 0e8c46ae08 Run isort and black on test files hai 1 ano
losses d8aacc510c return z_loss (#768) hai 1 ano
models 73df3be7d5 Add test for BTLM init hai 1 ano
modules 0e8c46ae08 Run isort and black on test files hai 1 ano
ops f5b308e258 [LayerNorm] Rename layernorm.py -> layer_norm.py hai 1 ano
pyproject.toml 73bd3f3bbb Move pyproject.toml to flash-attn and tests dir to avoid PEP 517 hai 1 ano
test_flash_attn.py 299563626f Fix test with alibi and cache_leftpad hai 6 meses
test_flash_attn_ck.py d8f104e97a Support AMD ROCm on FlashAttention 2 (#1010) hai 6 meses
test_rotary.py f692b98d80 Fix spurious re-compilations of `rotary_kernel` (#911) hai 9 meses
test_util.py 496fdc4f6c Add seqused_q in fwd / bwd and seqused_k in bwd. hai 4 meses