5 people 29468dde8b [Bugfix][KV Cache][MLA] Align packed block strides for V3.2 sparse MLA (#55528) ...

Signed-off-by: lz <145014769+200lz@users.noreply.github.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: OpenAI Codex <codex@openai.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>

2026-09-25 05:33:36 +00:00

..

test_attention_backends_selection.py

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

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

test_attention_backends.py

[ROCm][Bugfix] Ignore descales for unquantized AITER caches (#56726)

2026-09-17 01:29:50 +00:00

test_attention_splitting.py

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

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

test_b12x.py

[Kernel] Add B12X causal paged attention backend (#52017)

2026-09-01 10:42:02 -04:00

test_backend_per_kind.py

[Kernel] Resubmit PR 48666 - Gemma4 FP8 KV FA4 head dim 512 backend selection (#53175)

2026-09-23 20:29:06 -07:00

test_batch_reordering.py

[Attention] Support distinguishing between short extends and decodes (#37303)

2026-03-20 10:49:36 -07:00

test_chunked_local_attention.py

[Deprecation] Deprecate items scheduled for 0.29 (#55353)

2026-09-11 08:17:02 -07:00

test_cpu_mla_backend.py

[CPU][MLA] Fix prefill backend selection so MLA runs end-to-end on CPU (#51471)

2026-08-28 12:36:55 +08:00

test_cuda_backend_probe_errors.py

[Model] add GLM-5.3-Flash support (#53906)

2026-09-03 09:40:35 -07:00

test_dcp_a2a_pack_mask.py

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

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

test_deepseek_v4_rocm_adaptive.py

[ROCm][DSv4] Enable DSpark adaptive verification (#52362)

2026-09-21 10:03:41 -05:00

test_deepseek_v4_swa_visible.py

[Feat][Model] Support encoder-side SWA-bounded replay for DeepSeek-V4.1-Flash (#56227)

2026-09-18 15:01:24 -07:00

test_dspark_noncausal_sparse_mla.py

[Feat][Model] Support encoder-side SWA-bounded replay for DeepSeek-V4.1-Flash (#56227)

2026-09-18 15:01:24 -07:00

test_flashinfer_dcp_spec_reorder.py

[Attention] Extend XQA decode support on SM90 (#50439)

2026-09-11 06:19:55 +00:00

test_flashinfer_mla_dcp.py

[Spec decode] Support variable-length decode for Kimi-K3 adaptive ver (#52988)

2026-09-22 20:43:32 -07:00

test_flashinfer_mla_sparse_sm90.py

[Dependency] Upgrade FlashInfer version to 0.7.0 (#58069)

2026-09-23 22:45:23 -07:00

test_flashinfer_sparse_mla_sm120_api.py

[Bugfix][SM120][MLA] Disable dense prefill for FlashInfer sparse MLA (#51395)

2026-08-17 16:58:02 -07:00

test_flashinfer_sparse_mla_workspace.py

[Bugfix][Attention] Stabilize sparse-MLA DCP for GLM PCP evals (#55879)

2026-09-15 09:58:30 -07:00

test_flashmla_nope_sm90_backend_selection.py

[perf] wire FA and FlashMLA for sm90 GLM5Next NoPE SparseMLA (#55385)

2026-09-22 02:03:59 -07:00

test_gdn_metadata_builder.py

[Bugfix][GDN] Fix stateless first-chunk classification (#51565)

2026-09-22 09:11:49 +08:00

test_group_head_counts.py

[Bugfix][CPU] Fix several bugs (#54042)

2026-08-31 15:33:06 +08:00

test_group_sliding_window.py

[Bugfix] Take the sliding window from the layer, not the KV cache group (#51756)

2026-08-11 09:10:46 -07:00

test_indexer_dcp_localize.py

[Perf][Attention] Reduce GLM sparse MLA preparation overhead (#57458)

2026-09-21 17:24:25 -07:00

test_indexer_deepseek_v4_slot_mapping.py

[Feat][Model] Support encoder-side SWA-bounded replay for DeepSeek-V4.1-Flash (#56227)

2026-09-18 15:01:24 -07:00

test_indexer_native_next_n.py

[Attention][DSA] Take the native decode path for MTP=3 on SM90 (#52164)

2026-08-14 10:09:51 +00:00

test_kpool_tail_slot_mapping.py

[Perf][GLM] Fuse the kpool tail slot mapping into one Triton kernel (#57534)

2026-09-20 18:20:55 +08:00

test_kv_head_stride_canonicalization.py

[Attention] Make FlexAttention and FlashAttention use num-blocks first layouts (#42095)

2026-05-26 19:55:56 -07:00

test_linear_attention_metadata_builder.py

[Feature] Support MTP speculative decoding for Bailing hybrid models (#44880)

2026-07-06 10:38:50 +08:00

test_mamba_update_block_table.py

[Perf] Batch Mamba2 prefill SSM state saves, removing GPU<->CPU syncs (#49371)

2026-09-25 01:11:19 +00:00

test_mla_backends.py

[Bugfix][KV Cache][MLA] Align packed block strides for V3.2 sparse MLA (#55528)

2026-09-25 05:33:36 +00:00

test_mla_context_chunks.py

[3/N] HiSparse: host-resident sparse-MLA decode hot-buffering (#53781)

2026-09-12 00:36:47 -04:00

test_mla_noncausal.py

[Bugfix][MLA] Restore DSpark cache-group capability under optimized Python (#55234)

2026-09-04 10:24:31 -07:00

test_mla_prefill_quant_output.py

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

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

test_mla_prefill_registry.py

[Attention][MLA] Fuse Kimi-K3 chunked-context K/V packing (#51772)

2026-08-12 22:42:57 -07:00

test_mla_prefill_selector.py

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

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

test_mm_prefix.py

[Attention] Add Triton/FlashInfer composite for multimodal prefix attention (#56305)

2026-09-13 14:29:41 +00:00

test_replayssm_metadata_builder.py

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

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

test_rocm_aiter_mla_fp8_decode_routing.py

[CI][ROCM] Add the Fusion E2E TP2 Quick group on MI355, and the AITER MLA fix it needs (#58369)

2026-09-23 21:01:58 -05:00

test_rocm_aiter_mla_mtp_split.py

[CI][ROCM] Add the Fusion E2E TP2 Quick group on MI355, and the AITER MLA fix it needs (#58369)

2026-09-23 21:01:58 -05:00

test_rocm_attention_backends_selection.py

[ROCm] Give turboquant boundary layers a layout-compatible backend (#54988)

2026-09-23 22:00:07 -05:00

test_rocm_glm5next_sparse.py

[ROCm][Bugfix] Route GLM-5.3-Flash MTP through ragged sparse MLA (#55239)

2026-09-10 20:13:14 -05:00

test_sparse_indexer_decode_seq_lens.py

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

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

test_sparse_mla_backends.py

[Perf][Attention] Reduce GLM sparse MLA preparation overhead (#57458)

2026-09-21 17:24:25 -07:00

test_sparse_mla_kv_cache_layout.py

[Bugfix][KV Cache][MLA] Align packed block strides for V3.2 sparse MLA (#55528)

2026-09-25 05:33:36 +00:00

test_sparse_mla_mask.py

[Attention] Vectorize sparse MLA mask loads (#52217)

2026-08-18 20:32:41 -07:00

test_trtllm_attention_integration.py

[Perf][Attention] Bound FlashInfer prefill dequantization scratch (#57918)

2026-09-24 10:42:08 -04:00

utils.py

Add pydocstyle to the ruff rules (#52136)

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