Yongye ZhuandClaude Opus 5.5 a6009ea85b [Kernel][Build] Bump DeepGEMM to the 26/09/30 release and adapt the sparse indexer (#61149) ...

Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

2026-10-12 23:24:43 -07:00

..

flydsl_kernels

[Attention] Add UltraQuant 4-bit KV cache backend (FlyDSL D=256) (#57057)

2026-10-06 20:42:52 +00:00

turboquant_soa

[CI] Ignore ruff D209, rejoin the docstrings it split, and silence incompatible-rule warnings (#57212)

2026-09-16 10:57:24 -07:00

ultraquant

[Attention] Add UltraQuant 4-bit KV cache backend (FlyDSL D=256) (#57057)

2026-10-06 20:42:52 +00:00

__init__.py

[1/N][Attention] Restructure attention: move files (#31916)

2026-01-09 13:10:24 -08:00

chunked_prefill_paged_decode.py

[ROCm] [bugfix] Chunked prefill paged decode masked load perf (#50017)

2026-08-12 23:04:01 +00:00

common.py

[Kernel][Build] Bump DeepGEMM to the 26/09/30 release and adapt the sparse indexer (#61149)

2026-10-12 23:24:43 -07:00

cp_common.py

[refactor] consolidate cp attn ops (#52839)

2026-08-19 18:04:11 -07:00

dcp.py

[Spec Decode] Enable Gemma4 DSpark adaptive verification with FlashInfer (#57263)

2026-09-26 17:01:17 +08:00

flashmla.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

flydsl_turboquant_decode.py

[Kernel][ROCm][Perf] FlyDSL decode-attention kernel for 4-bit TurboQuant KV cache (#47896)

2026-08-11 08:45:39 -07:00

flydsl_ultraquant_decode.py

[Attention] Add UltraQuant 4-bit KV cache backend (FlyDSL D=256) (#57057)

2026-10-06 20:42:52 +00:00

int4_per_token_head.py

[Bugfix][Attention] Support int4_per_token_head on non-power-of-two head sizes (#56198)

2026-10-05 04:52:39 +00:00

merge_attn_states.py

[ROCm][Perf] Use native merge_attn_states kernel on gfx942/gfx950 (#60159)

2026-10-08 04:49:49 +00:00

metadata.py

[Spec Decode] Remove eager metadata rebuild during MTP fused multi-step decode (#58463)

2026-10-06 12:17:30 -04:00

paged_attn.py

[XPU][9/N] clean up existing ipex code/doc (#34111)

2026-02-11 00:27:15 -08:00

pcp.py

[Perf][PCP] Shard decode requests across PCP ranks (#52162)

2026-10-01 23:50:13 -04:00

prefix_prefill.py

[Kernel] Make prefix-prefill tiling independent of the KV page size (#54194)

2026-09-01 09:19:55 +08:00

rocm_aiter_mla_merge.py

[ROCm][MLA][DCP] Support causal multi-token verification (#51705)

2026-08-31 10:42:28 -07:00

rocm_aiter_mla_prefill.py

[ROCm][MLA] Keep DCP prefill context in FP8 through AllGather (#59070)

2026-10-13 09:20:38 +08:00

rocm_aiter_mla_sparse.py

[ROCm][Perf] Skip cleaning sparse prefill MQA logits (#51314)

2026-10-13 00:21:52 -05:00

rocm_paged_mxfp4_indexer.py

[ROCm][DSv4.1] Paged MXFP4 sparse indexer on aiter's MQA-logits kernel (#58671)

2026-09-29 14:37:57 +08:00

segmented_attention.py

[ROCM][Perf] Add ROCM_SEGMENTED_ATTN attention-backend works across RDNA3/3.5/4 (#59132)

2026-10-06 07:53:47 +00:00

triton_attention_helpers.py

[Attention][Spec Decode] Multi-query Triton split-K for spec-decode verify, with empty-segment/padding hardening (#56148)

2026-10-13 04:37:56 +04:00

triton_decode_attention.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00

triton_merge_attn_states.py

[Attention][MLA] Per-request scheduling for MLA chunked context (#50613)

2026-08-06 16:45:44 +00:00

triton_prefill_attention.py

[Model] Support EmbeddingGemma2 multimodal pooling architecture (#60254)

2026-10-06 13:43:38 -04:00

triton_reshape_and_cache_flash.py

[2/N][KV-Cache Layout Refactor] Pack K/V into the content dim across attention backends (#44455)

2026-07-11 11:11:16 -04:00

triton_turboquant_decode.py

[Attention][TurboQuant] Share dequant buffers, eliminate float16_copy (#40941)

2026-04-27 13:48:36 +08:00

triton_turboquant_store.py

[2/N][KV-Cache Layout Refactor] Pack K/V into the content dim across attention backends (#44455)

2026-07-11 11:11:16 -04:00

triton_unified_attention_diffkv.py

[Feature][Attention] Cover a whole spec-verify in one DiffKV split-KV program (#59085)

2026-10-10 11:55:45 +08:00

triton_unified_attention.py

[Attention][Spec Decode] Multi-query Triton split-K for spec-decode verify, with empty-segment/padding hardening (#56148)

2026-10-13 04:37:56 +04:00

vit_attn_wrappers.py

[MM] Fix compiled ViT attention output layouts (#58182)

2026-09-29 06:27:03 -07:00

xpu_mla_sparse.py

Add pydocstyle to the ruff rules (#52136)

2026-09-16 16:44:29 +00:00