-
Notifications
You must be signed in to change notification settings - Fork 461
Open
Labels
bugSomething isn't workingSomething isn't working
Description
Your current environment
2025-09-19T19:13:52.5656099Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5657400Z a22b532d38fe6f1cda1a2e61d1f9af8888cac256 [Fixbug] Fix shape not match when sliding_window and dynamic batch_size (#2830)
2025-09-19T19:13:52.5658591Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5659868Z 0942d9aaabb6344b542f43147add9077e9c49c3c [3/N][Refactor][Quantization]remove packed_modules_mapping from models (#3021)
2025-09-19T19:13:52.5661057Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5662178Z 833cd1b698f3d467bb0a6a60cbf20ebc5535f5c9 [BugFix] Async scheduling and PP compatibility with DP (#2796)
2025-09-19T19:13:52.5663277Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5664350Z 0a526768f55cc60b5b870adce92affdfcfea8523 [Feature] Support moe multi-stream for aclgraph. (#2946)
2025-09-19T19:13:52.5665410Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5666870Z 6681dde9028b2ae8ab0e56158a117c5db7f0bc7a [Feat][Graph] Support MTP for ACL Graph (#2932)
2025-09-19T19:13:52.5667935Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5668972Z cef43b524e5dbf24434ac330235c5c835284c580 [Feat] A Connector that supports Mooncake store (#2913)
2025-09-19T19:13:52.5670003Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5671072Z 76844eec78a23f482a4e0dfe9684898a6ef35fb2 Dynamic Expert Load Balance with Zero-like-overhead (#2956)
2025-09-19T19:13:52.5672130Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5673230Z ae758dda05b57adaf70af44122e6bdc54fbb88ff [Bugfix] Fix mtp torchair in pd Disaggregation scenario (#2951)
2025-09-19T19:13:52.5674313Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5675395Z 6b7117dbb74e6c46da110d74014af811be6323ec [main] addrmsnorm + quant fusion optim in Dense Models (#2772)
2025-09-19T19:13:52.5676452Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5677465Z 88ca8a051ca51fe72516344db092a7852150cfdb [Feat][Graph] Support DeepSeek with ACL Graph (#2707)
2025-09-19T19:13:52.5678474Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5679463Z 1c5900327b67015e5707d3879ccf5fa5ab622832 [refactor] refactor deepseek-related files (#2849)
2025-09-19T19:13:52.5680451Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5681850Z 18ca7861f6e4f27cea58cc1d70a8c3081422081c [Main] [Refactor] Enable MoECommMethod in Eager Mode (#2791)
2025-09-19T19:13:52.5682943Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5683894Z c556038ef0b8580cb9079823bd9f06263ee7d731 [New model] Qwen3-next support (#2917)
2025-09-19T19:13:52.5684800Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5685822Z 382c29f3e1a3201c5bedd588f50fbe55dad2d919 [BugFix] Fix world size bug in model_runner (#2915)
2025-09-19T19:13:52.5686821Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5687866Z c5a502fd2e81e6dfc0cbafbc4ffa49bb68c3abf1 main add ascend scheduler support multimodal (#2844)
2025-09-19T19:13:52.5688898Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5689904Z 0a27705917e64993a8a76198ac6e30980578fe60 fix mooncake connector adxl hostname usage (#2824)
2025-09-19T19:13:52.5690910Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5692042Z e57cca971c4db11d8ce9da6b008bd655ada8c77e Fix the bugs about operator registration by PyTorch Dispatcher (#2786)
2025-09-19T19:13:52.5693158Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5694252Z 585a494baa4bdbce5a71ef6808033466ed9f90f3 [Core] Disable the chunked prefill feature in Non-MLA LLMs (#2894)
2025-09-19T19:13:52.5695334Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5696390Z 756b8a1946aa9396d5bc7b9c67547fcb93fad630 Revert "[Feat] Unquantized linear nz support (#2619)" (#2896)
2025-09-19T19:13:52.5697430Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5698525Z fc2bcbe21c86f7684c80e42771b128da9fc17571 [Ops] Fix bug in register_custom_ops without forward_context (#2883)
2025-09-19T19:13:52.5699817Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5701041Z 778cb7255697ad0d1562f60d0cf4ef68542a5b94 fix bug when rotary_dim is not 128 (#2847)
2025-09-19T19:13:52.5702001Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5703121Z f5a97e8fa5440df6735d1121f813cda7f1257367 [Quantization] register AscendQuantRMSNorm for quantization (#2856)
2025-09-19T19:13:52.5704261Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5705513Z eab3635850ba351af81d76a7b4b3db46ffb7f697 [Bugfix] Retrieve num_redundant_experts from eplb_config in torchair qwen3_moe.py (#2857)
2025-09-19T19:13:52.5706754Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5707856Z aeffe27b3089cf12b743806fd922c7d1fd455ac2 [Perf]set moe w2_weight default to be nz (#2842)
2025-09-19T19:13:52.5708873Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5709850Z 9615dea3a71df8ecd2c591f284d9615140dce68a Refactor tensor_parallel and comm_utils (#2814)
2025-09-19T19:13:52.5710842Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5711862Z 0005479b9c3ebf262d379a454641d407a8a1dba6 [main] mlp weight prefetch in Qwen Dense Models (#2816)
2025-09-19T19:13:52.5712885Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5713906Z c3c222150363900e5f0e87ae2abb4868dacf6a1c [Feat]support dynamic quantization in allgather (#2841)
2025-09-19T19:13:52.5714930Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5716123Z bd3dedea6123c9c8c19fe83b6e05716f63b1285d support qwen25 vl w8a8 quantization (#2778)
2025-09-19T19:13:52.5717136Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5718104Z 2b9269b581e390357089d54fb8934ea65bf0ced4 [Perf][V1] Fully overlap model execution (#2783)
2025-09-19T19:13:52.5719079Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5720043Z 923cdaeba389ea5b895acb25a7c755ccb830cee2 fix ascend fused moe spelling error (#2863)
2025-09-19T19:13:52.5720998Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5721995Z b9a0a75c783571caf22129612fb3338272d1782c fix qwen torchair attention PrefillCacheHit (#2787)
2025-09-19T19:13:52.5722991Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5723962Z 7b2ecc1e9a64aeda78e2137aa06abdbf2890c000 [Feat] Unquantized linear nz support (#2619)
2025-09-19T19:13:52.5725180Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5726291Z 5691104249bbee7648e8cfc1466a96c092a8d76d LLMdatadist connector adapt the distributed KV aggregation (#2718)
2025-09-19T19:13:52.5727382Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5728430Z c2fdd4b8bc9ce859343909caebf331e0cd047908 [CI/UT] Fix UTs on register customop and warm up model (#2862)
2025-09-19T19:13:52.5729476Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5730469Z aa4d2a91ed6450759895ee7fd614eaf336c4722e Refactor AscendMultiHeadLatentAttention (#2826)
2025-09-19T19:13:52.5731459Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5732461Z 168ad600b5d794fef4314980ddeac9f71511c449 [main] add pd transfer for ascend scheduler (#2753)
2025-09-19T19:13:52.5733461Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5734835Z edf1f600ad30120a7d870a38b65520616f3131ae [CI] Remove compatibility maintenance for vllm v0.10.1 and v0.10.1.1 (#2840)
2025-09-19T19:13:52.5735959Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5737033Z c735bb09419beb0fa9a186ce06cc9ddf2c3cc50b [Fix] Ensure metadata sync across DP ranks in eager mode (#2766)
2025-09-19T19:13:52.5738107Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5738965Z 2693196ef8382761ae7858b9fcafe3866bf7287a add gatherep select. (#2740)
2025-09-19T19:13:52.5739810Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5740760Z 6666e5265d40ecafc3cb377233fee840d7fe553b Added support for KV connector v1 (#2039)
2025-09-19T19:13:52.5741702Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5742579Z f86596a66cc0aff2b05280303212758380f0ec9a allgather use fusedop. (#2689)
2025-09-19T19:13:52.5743433Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5744660Z 7d47d8f4f61632b52a35068eadc91d1934e0b71b [Fix] fix resources limit error when apply speculative decoding and aclgraph (#2472)
2025-09-19T19:13:52.5745858Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5746860Z 0c0789be7442122eb1203abbf89a9592648922e0 [Feat] allow using aclgraph in ray backend (#2589)
2025-09-19T19:13:52.5747979Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5749453Z aff5189c8781f1127d7dbca6c59a9ec44a427100 [main] Fuse GroupedMatmul, Swiglu and DynamicQuant in `W8A8_DYNAMIC` quantized MoE layers (#2275)
2025-09-19T19:13:52.5750757Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5751872Z 37f5a29cd4f84fa0beee236dadf070b41e4b5403 [1/N][Refactor][Quantization] remove redundant quantizer class (#2680)
2025-09-19T19:13:52.5752973Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5753915Z d4370ebc42f8a2cecbb7ad4b199ac3f840ca3b28 [Refactor] Refactor Spec Decode (#2668)
2025-09-19T19:13:52.5754847Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5755927Z e7409e95ee73fb3bb7bf8b23f26c16620ed94543 [1/N][Draft][Refactor]torchair pangu_moe modeling refactor (#2437)
2025-09-19T19:13:52.5756990Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5757997Z a58013440a9c9c0b5220a60bc161025c5f5270a2 [BugFix][MLA] Fix attn_mask bug for ring mla (#2704)
2025-09-19T19:13:52.5758989Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5760146Z 984bd7c13a6b7eb80ac9cb43ab85a81afe779614 [Bugfix][APC] Fix accuracy issue on prefix caching with AscendScheduler (#2714)
2025-09-19T19:13:52.5761285Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5762271Z df88a2ecc8116a42d79a13fa1a8a05a03c70324f [P/D]mooncake_connector adapted to 0.10.1 (#2664)
2025-09-19T19:13:52.5763250Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5764308Z 07d44ade194b018ae2cc172482d55cb746c5fd0e bugfix: fix initialization error for mooncake in k8s (#2541)
2025-09-19T19:13:52.5765346Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5766613Z 90a75a90a9adc8be8efa294bada8391bb1607e0d [bugfix] fix torchair runtime error caused by configuration mismtaches and file missing (#2532)
2025-09-19T19:13:52.5768079Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5769496Z 5889fa1b1cddb283b5e2c206d08062bef8432cbd [bugfix] ascend schedule encountered an incorrect req block length in the check_watermark_for_prefill function (#2508)
2025-09-19T19:13:52.5770886Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5771853Z 3584306387bc5d094700e71c75fbc9b5154bdaf7 [Bugfix] Fix qwen2.5-vl-without-padding (#2623)
2025-09-19T19:13:52.5772827Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5774122Z eaeb2efb20ff70875991483d63262f379a3afde8 [Main][Feat]Set the Profiler parameters through environment variables consistent with vLLM (#2608)
2025-09-19T19:13:52.5775414Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5776483Z 93754d80616830a5bc068c51d3493b84f679750d [Bugfix] Fix long context seq accuracy problem for `GLM4.5` (#2601)
2025-09-19T19:13:52.5777558Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5778666Z b84465c52564134d217a0119e34a68ef86b6d35d [Perf]Enable npu_moe_gating_top_k_softmax on quantized scenarios (#2633)
2025-09-19T19:13:52.5779771Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5780801Z c1e607b7b71c7b93d2ac3b04b0cc3971da896473 [Misc] Clean up uesless code in rotary_embedding (#2663)
2025-09-19T19:13:52.5781817Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5782778Z 253b01b9a552b9745aaea827bd8944c87ea7b09a [7/N][refactor]fix torchair rope ops (#2683)
2025-09-19T19:13:52.5783948Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5785110Z 9f1e054fe3df966a4fa51cc73e34c430452b5ebc [Bugfix][LoRA][Operator] Fix LoRA custom operators accuracy issue (#2672)
2025-09-19T19:13:52.5786244Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5787336Z 214b32a3460809148271d66774a457e3c750d79e [V1][BUGFIX][0.10.1] FIX mtp on main branch (#2632)
2025-09-19T19:13:52.5788316Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5789498Z 0df059f41a4c45ac98b0b6be9d5b50f989893905 [CI] Fix CI Break: upstream adds routed_scaling_factor in forward_oot interface (#2675)
2025-09-19T19:13:52.5790692Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5791570Z ea53f9076e722eb669d9df76ed6601d807acae7e support torchair mode (#2641)
2025-09-19T19:13:52.5792433Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5793438Z ad13964c7121d7d80813c6f79a0b5fce9b6f66b0 [6/N][refactor]delete torchair in rotary ops (#2581)
2025-09-19T19:13:52.5794433Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5795403Z c2c97f3079957efafbfba29ba97ded417e23fd55 [5/N][refactor]add torchair rotary ops (#2559)
2025-09-19T19:13:52.5796372Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5797502Z 3a5fc5ee01edb3e6c88774edfea2a85eed1ff990 [Refactor][MoE] remove redundant code after refactoring fused_moe (#2612)
2025-09-19T19:13:52.5798642Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5799534Z 20ae71291d876a8511eb504601f747a60864861d [torchair]remove aicpu op (#2640)
2025-09-19T19:13:52.5800424Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5801521Z 7215454de6df78f4f9a49a99c5739f8bb360f5bc bugfix for torchair graph (#2639)
2025-09-19T19:13:52.5802404Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5803576Z d3c93fba5ca9279ec9f0ebb9026f90abf6132f38 [3/N][Feat][Graph] Support `all-to-all` and quantized models with ACL Graph (#2614)
2025-09-19T19:13:52.5804748Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5805828Z 91c35d765aa2edeb3e9c805f2fe3c330320fe696 [Bugfix] Fix mc2 operator error in aclgraph + ep<16 scenario (#2609)
2025-09-19T19:13:52.5806913Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5808012Z 52aff9e229b8c5557bb2ee485ec381e0886d3701 [main] [bugfix] Fix misjudging quantized/unquantized scenarios (#2627)
2025-09-19T19:13:52.5809104Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5810212Z aadc75c247924ab8e90c3d82f0ccabcc48cf90ab [Fix] Resolve data-parallel (DP) assertion errors in TorchAir (#2626)
2025-09-19T19:13:52.5811323Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5812337Z 600b08f7542be3409c2c70927c91471e8de33d03 [Feat]: Add custom lmhead tensor model parallel (#2309)
2025-09-19T19:13:52.5813361Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5814308Z dfc7eb39ada3f86f5c15425ba759ecfaa8f5c9a8 [Fix] Fix DP-related padding logic (#2582)
2025-09-19T19:13:52.5815259Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5816231Z 175f6bc445173704ad8f2a747a08866a205c4b39 Support v0.10.1 (#2584)
2025-09-19T19:13:52.5817080Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5818093Z 6c973361fc2eba5d3faa9b6b496b4b9fec4dc784 [Bugfix] Fix aclgraph not enabled by default (#2590)
2025-09-19T19:13:52.5819096Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5820331Z cf96366a396b7d70b390cb244b53c58e9666c4d5 [Bugfix][LoRA][Patch] Fix the LoRA inference bug after upstream vLLM codebase changed (#2560)
2025-09-19T19:13:52.5821557Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5822944Z 1191a64ae508183d5613711bc98a90250963f83a [Feat]attention add sliding windows size (#2528)
2025-09-19T19:13:52.5824083Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5825392Z c8d1df3a3fa803a8a0742df80cf7601a8b15ef7a [Refactor][WIP] Refactor mla_v1 by moving all MLA preprocessing ops into mla_v1 attention impl (#2465)
2025-09-19T19:13:52.5826682Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5827905Z 320edde2df14a16436d7094012c12592b2e16266 [main] [refactor] refactor fused_moe.py to enable token_dispatchers (#2570)
2025-09-19T19:13:52.5829037Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5829980Z 936c102105b72a4e36dd284f900f32338a232696 [bugfix][refactor]fix torchair_w8a8 (#2569)
2025-09-19T19:13:52.5830916Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5832164Z a955e5d4046e4e3a55976e88774073b3c56b463b [4/N][refactor]delete torchair from quantization (#2535)
2025-09-19T19:13:52.5833514Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5834708Z c578f817ca4c17a076ac7fa93de77db11f008fae [CustomOp] Register VocabParallelEmbedding instead of overwrite forward (#2515)
2025-09-19T19:13:52.5836179Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5837297Z 2bfbf9b9b3f6cd332fc438fc322b00b6053a043c [main][bugfix] Fix bugs and refactor cached mask generation logic (#2442)
2025-09-19T19:13:52.5838420Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5839762Z 6881c194580b800cb233632300575b9b387778ef [main] convert the format of gmm to nz (#2474)
2025-09-19T19:13:52.5840864Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5841811Z 20a7bc4b71827a39c21ecac95536183468ad90a7 [3/N][refactor] refactoer quantization (#2504)
2025-09-19T19:13:52.5842764Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5843796Z acdc53c2f6b480f23f40ea9e72356182386e0bac [Bugfix] Fix the bug of cos invalid shape when dp (#2558)
2025-09-19T19:13:52.5844828Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5845953Z a9e78a329988c4fbe5382ef03b08d6413623d248 [Aclgraph] Update compilation config in `check_and_update_config` (#2540)
2025-09-19T19:13:52.5847106Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5848085Z f22077daa6a32e1d5c5cfe0e84da2cea1ab8cafb [Embedding] Recover embedding function (#2483)
2025-09-19T19:13:52.5849091Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5850082Z 6a4ec186e731b9516235f4fd30b5b98227513fe7 [Qwen-moe] Remove the minor operation arange (#2373)
2025-09-19T19:13:52.5851081Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5852322Z 358ba6899401e5a1f5e8860e8ea88900cc27dd38 [main][bugfix] Fix MatmulNZ format bug on some machines (#2549)
2025-09-19T19:13:52.5853411Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5854443Z a6bb502e70b7554b2a0342565348b1a191cd0aa0 [2/N][Feat] Add MC2 communication method for MoE layers (#2469)
2025-09-19T19:13:52.5855481Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5856489Z 5d8ec280090b4a7567fb2b50a7cedda44902c37f [2/N][refactor] split torchair from fused_moe (#2503)
2025-09-19T19:13:52.5857484Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5858647Z cfe77e83aeda343274c0488b93e2263bee44a860 [Bugfix]Support Qwen3-MOE on aclgraph mode in sizes capture and add new ut (#2511)
2025-09-19T19:13:52.5859843Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5861051Z b3fdd78a6b6f8fe9546a1c1092926293577c1b50 [Main][Refactor]Change ASCEND_QUATIZATION_METHOD to ASCEND_QUANTIZATION_METHOD (#2517)
2025-09-19T19:13:52.5862264Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5863098Z 7e494e94a969626d07d2414cb1fb859d2206e3b8 [CI] Fix broken ci (#2530)
2025-09-19T19:13:52.5863968Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5865346Z 99bf25af76a9f6759e262e351c7d41fed57f159f [Fix] Add operations in `_dummy_run` to maintain synchronization with `_process_reqs`, resolving a service hang (#2454)
2025-09-19T19:13:52.5866700Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5867907Z de7649492ddcbdb7c818665f0b81cc8fbaaaa4b7 [Refactor] cleanup converting_weight_acl_format_format (#2482)
2025-09-19T19:13:52.5868987Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5870277Z 0f81e032f04b72f4dd0c7fefd62b7220942c545a [1/N][refactor] torchair fused_moe refactor (#2438)
2025-09-19T19:13:52.5871252Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5872401Z f796e6280b79cc87451ceac2daa086ee80b1d572 [CustomOp] Register RotaryEmbedding instead of overwrite forward (#2385)
2025-09-19T19:13:52.5873539Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5874513Z 950c4b219a8fe5d4339de3acd535b32b99e790ad [main] refactor alltoallv in fused_moe (#2487)
2025-09-19T19:13:52.5875460Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5876728Z 4af5b80606e6cffe440c27237655cd44c2e5bdaf [Scheduler] validate max_num_batched_tokens and max_model_len in AscendSchedulerConfig (#2434)
2025-09-19T19:13:52.5877979Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5878926Z 3629bc4431d3edb4224761f9036b3bddb16158d6 feat: add mtp ut and fix some bugs (#2453)
2025-09-19T19:13:52.5879843Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5880846Z dd04a96ee3caa8c85fbc72a6328b969d1c373bc9 [Bugfix] Fix the bug of incorrect precision (#2479)
2025-09-19T19:13:52.5881869Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5882674Z b0403f8d8a5ced0ddb86722754bc515f4df9d1f1 [CI] fix ci (#2464)
2025-09-19T19:13:52.5883465Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5884500Z 0ca3f48c900b333673830e8307c259acc684c1a3 [2/N][refactor] torchair deepseek mla backend refactor (#2459)
2025-09-19T19:13:52.5885736Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5886644Z 3fb80ee356781752ed94ea7a39953bdd47eac764 add mlp tp optimze (#2120)
2025-09-19T19:13:52.5887489Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5888372Z 0dca4c6dbdbe4b77116fea6219f6ca494f63e9d2 refact runner model v1 (#2461)
2025-09-19T19:13:52.5889249Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5890445Z 1de16ead8eecfec8903ec1b330b27a4fa2593c35 [main][bugfix] Modify the default value of the enable_shared_pert_dp to false (#2457)
2025-09-19T19:13:52.5891632Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5892740Z c40d4171bcc0424ad88fc8c9bdd6f694170a8342 [main][quantization] Adapt to the new format of ds w4a8 weight (#2392)
2025-09-19T19:13:52.5893856Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5894880Z 3f867ee7081f6f041652180ccf06d0c70fd44429 refactor allgather/mc2-related fused_experts (#2369)
2025-09-19T19:13:52.5895878Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5896829Z 73acdcfc3bb56363a7ae52cd2af0d0cf84f55592 [PD] Correct the ip and port env (#2450)
2025-09-19T19:13:52.5897777Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5898750Z 7bec1a9b9c372785551d45682bf11063ec42b216 qwen3_moe/qwen25 support torchair graph (#2403)
2025-09-19T19:13:52.5899713Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5900628Z 31ae2497425cf26fd8aaedd9845b6066cd06fb84 [misc] remove uesless envs (#2448)
2025-09-19T19:13:52.5901534Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5902533Z 1327f9be1cea85b2750ac4145981ddd732064bb9 Fix some ci issue and refactor modelrunner (#2445)
2025-09-19T19:13:52.5903774Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5904856Z d91c6daf891158d11113767bd85c0fe1eae2cde9 [improve] Remove redundant parentheses in pangu_moe.py (#2081)
2025-09-19T19:13:52.5905929Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5907013Z 83e0f41408fb92b7384bed8fdd1239fa6cf18b0b [3/N][Refactor] Move `torchair_attention` to `torchair` dir (#2017)
2025-09-19T19:13:52.5908192Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5909138Z 3f4a358b140226e5c6d218742ff10a210cda5800 [Bugfix] Fix custom op register issue (#2409)
2025-09-19T19:13:52.5910123Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5911145Z 3648d18e673f15a33a82d6ea95d3a9dd891ff1f5 Add Custom Kernels For LoRA Performance (#2325)
2025-09-19T19:13:52.5912136Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5913222Z 3fc31ee1cbdf0c0d11efc4da5fd865bb2d077c4b [1/N][refactor] torchair deepseek modeling refactor (#2384)
2025-09-19T19:13:52.5914281Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5915281Z 03ca2b26ca9ab6b9a12f021b0595a726ee35e223 [P/D] Mooncake Connector for v1 distributed (#1568)
2025-09-19T19:13:52.5916277Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5917317Z 2bb7e55022c3a558145a1b17ba3c93b4ab6bf00f [Bugfix][PD]fix non-working disaggregated prefill (#2374)
2025-09-19T19:13:52.5918537Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5919520Z 1b40665548f048c3417b79207bb6f1a930475624 [Misc] remove unused file (cache.py) (#2377)
2025-09-19T19:13:52.5920461Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5921592Z 61866b8ac6e8812a205a45dd1f4baee8919df028 [Quickfix] update CachedRequestState as NewRequestData changed (#2367)
2025-09-19T19:13:52.5922722Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5923826Z 2f50304c19bd7bac8fece88bcb4b273d6e64b412 [Bugfix] Add get_supported_tasks interface to fix broken CI (#2023)
2025-09-19T19:13:52.5924904Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5925787Z c59d69d9e65de4b91628411ef415eca6bf512b44 [PERF]support MERRouter (#1421)
2025-09-19T19:13:52.5926664Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5927779Z 8fa188111da3a8f752dc309330d9bd3ec18194e6 [PERF]support H2P communication optimization for PanguProMoe (#1463)
2025-09-19T19:13:52.5928882Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5930075Z 5c53cbaf2a7efcd09f2860eb6dff7412c29654b8 [BugFix]Fix bugs when initializing communication groups with dp on 300I Duo (#1478)
2025-09-19T19:13:52.5931257Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5932435Z 5f4391652f4a62d791fa4b9dfa3fc9d802d5a250 [PromptLogprobs][V1] Support prompt logprobs to fix ceval accuracy in V1 (#1483)
2025-09-19T19:13:52.5933619Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5934775Z d59e7fa0959a5571e7debd884d27ea2e6d9cb582 [CI] Pin transformers<4.53.0 and fix EPLB load_weights to make CI passed (#1482)
2025-09-19T19:13:52.5935928Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5937041Z 5968dff4e000f8c4b00751d963898f1f0619b164 [Build] Add build info (#1386)
2025-09-19T19:13:52.5937913Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5938963Z 53c2d58ae18d4268024ceb3ba029e1114b36733c Handle with_prefill_across_dp for multistream mla (#1322)
2025-09-19T19:13:52.5940006Z ----------------------------------------------------------------------------------------------------
2025-09-19T19:13:52.5941191Z 2690697caa47ab8daee4083778020ea7c13c16c7 [Bugfix] Reset all unused positions to prevent out-of-bounds in GatherV3 (#1416)
2025-09-19T19:13:52.5942363Z ----------------------------------------------------------------------------------------------------
🐛 Describe the bug
It seems include many old commits need to be removed.
Potabk
Metadata
Metadata
Assignees
Labels
bugSomething isn't workingSomething isn't working