Lucas Wilkinson db56a59970

[BugFix] Fix FA3 IMA with FULL_AND_PIECEWISE and cascade attention (default) (#28702)

2025-11-14 12:19:22 +00:00

..

mla

[bugfix] correct local_chunk_len for DCP in reorg_kvcache with long context (#28526)

2025-11-13 11:29:22 -08:00

__init__.py

…

cpu_attn.py

[CI Failure] Fix backend selection for encoder-only models (#28534)

2025-11-13 10:11:27 -05:00

flash_attn.py

[BugFix] Fix FA3 IMA with FULL_AND_PIECEWISE and cascade attention (default) (#28702)

2025-11-14 12:19:22 +00:00

flashinfer.py

[DCP] Support Decode Context Parallel (DCP) for GQA with Flashinfer (#25438)

2025-11-14 11:24:10 +00:00

flex_attention.py

[CI Failure] Fix backend selection for encoder-only models (#28534)

2025-11-13 10:11:27 -05:00

gdn_attn.py

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

linear_attn.py

Convert formatting to use ruff instead of yapf + isort (#26247)

2025-10-05 07:06:22 -07:00

mamba1_attn.py

Remove deprecated fields from CompilationConfig (#27593)

2025-11-12 16:10:28 +00:00

mamba2_attn.py

Remove deprecated fields from CompilationConfig (#27593)

2025-11-12 16:10:28 +00:00

mamba_attn.py

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

pallas.py

[V0 deprecation] Remove no longer used get_metadata_cls (#28370)

2025-11-10 14:32:09 +08:00

rocm_aiter_fa.py

[ROCm][BugFix]Fix get_cu_count in rocm_aiter_fa.py (#28618)

2025-11-13 14:18:20 +00:00

rocm_aiter_unified_attn.py

[V0 deprecation] Remove no longer used get_metadata_cls (#28370)

2025-11-10 14:32:09 +08:00

rocm_attn.py

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

short_conv_attn.py

Remove deprecated fields from CompilationConfig (#27593)

2025-11-12 16:10:28 +00:00

tree_attn.py

[Attention] Refactor CUDA attention backend selection logic (#24794)

2025-11-11 07:40:44 -05:00

triton_attn.py

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

2025-11-13 01:56:40 +09:00

utils.py

Fix KV sharing fast prefill with cudagraph enabled (#28537)

2025-11-14 11:53:42 +00:00

xformers.py

[Attention] Refactor CUDA attention backend selection logic (#24794)

2025-11-11 07:40:44 -05:00