Commit Graph

Select branches

Hide Pull Requests

cch/megamoe-rc-v0.1

Mono Color

• 98224de29b [Bugfix] Add missing auto_create_handle_loop to communicator methods (#19610) Kangyan-Zhou 2026-03-01 15:00:05 -08:00
• 0b3ddbcf10 [fix typo] seperated_timestep -> separated_timestep (#19622) SoluMilken 2026-03-02 06:09:51 +08:00
• dc02e5bea7 [HiCache] Re-land spec v2 + decode KV cache offloading compatibility (#19615) Kangyan-Zhou 2026-03-01 13:58:31 -08:00
• 0e86977811 [RL] Support per-layer mixed FP8/BF16 serving for FP8 checkpoints (#18742) Ziang Li 2026-03-01 05:59:22 -08:00
• a75840b373 [diffusion] CI: create and refactor UT (#19619) Mick 2026-03-01 19:38:20 +08:00
• 80a6b32703 [Perf] Optimize NSA backend metadata under MTP (#19536) Brayden Zhong 2026-03-01 04:59:26 -05:00
• d098c8dab0 [diffusion] add .claude and update contributing with attitude towards vibe-pr (#19511) Mick 2026-03-01 14:41:55 +08:00
• 5fa6633485 [AMD] Fix MoRI EP warmup hang by restoring deepep_mode=normal default (#19498) Bingxu Chen 2026-03-01 14:05:22 +08:00
• dcf462cfba Revert "[HiCache] Enable spec v2 + decode KV cache offloading compatibility" (#19613) Kangyan-Zhou 2026-02-28 21:54:32 -08:00
• 8167346609 [HiCache] Enable spec v2 + decode KV cache offloading compatibility (#19518) Kangyan-Zhou 2026-02-28 21:53:52 -08:00
• 894e887e4a [Blackwell] Make mxint4 flashinfer_trtllm moe gemm set by default on blackwell (#18136) Yi Zhong 2026-03-01 00:21:01 -05:00
• 38dc372dae [Bugfix] Fix KeyError: 'prompt_tokens' when streaming requests are aborted (#19514) Jun Liu 2026-03-01 13:21:35 +09:00
• 4ec450e97b [diffusion][MOVA] fix: fix task type in MOVA pipeline and shared model placement (#19489) Zhang Yiyang (SII) 2026-03-01 12:13:15 +08:00
• 5acb45cf32 [Session] Extract SessionController and clean up session logic in Scheduler (#19547) Liangsheng Yin 2026-02-28 19:47:44 -08:00
• a45613f2a6 Revert "[SGL] sync patch: Remove sync points, prefill cudagraph for DP, disable cache reset in mem check (#19190)" (#19581) Alison Shao 2026-02-28 19:46:47 -08:00
• e64095c3c7 Support data parallel attention in dump comparator (#19602) fzyzcjy 2026-03-01 10:51:21 +08:00
• ea6ff7b01f Support multi sharding group on the same dimension in dump comparator (#19601) fzyzcjy 2026-03-01 10:36:48 +08:00
• 46960e65cf Add skip patterns, tee to file, tensor load warning in dump comparator (#19600) fzyzcjy 2026-03-01 10:36:22 +08:00
• b0b26a7ef1 Support concat mode in token aligner in dump comparator (#19599) fzyzcjy 2026-03-01 10:35:50 +08:00
• e78f1283f7 Support overriding and post-hoc providing metadata in dump comparator (#19598) fzyzcjy 2026-03-01 10:35:06 +08:00
• e41164af1c Enhance replicated tensor checker in dump comparator (#19597) fzyzcjy 2026-03-01 10:34:34 +08:00
• ec08240a6a Support data parallel in dump comparator (#19596) fzyzcjy 2026-03-01 10:34:03 +08:00
• 003ad6daaa Support partial tensors waiting for reduction and pipeline parallel in dump comparator (#19595) fzyzcjy 2026-03-01 10:33:39 +08:00
• 67810828cf Visualize per-token information in dump comparator (#19594) fzyzcjy 2026-03-01 10:32:59 +08:00
• f5a10e04cd Support arbitrary filtering in dumper (#19593) fzyzcjy 2026-03-01 10:31:21 +08:00
• 8240a87306 [AMD] MORI-EP support for EP4. (#19578) Duyi-Wang 2026-03-01 05:13:46 +08:00
• 560b867cce [AMD] Remove Redundant tvm-ffi Installation in amd_ci_install_dependency.sh (#19554) Bingxu Chen 2026-03-01 04:57:50 +08:00
• f451664504 [Fix] Add --disable-draft-model-update to control draft model updates(especially in RL) (#15726) Haodi Lei 2026-03-01 04:09:55 +08:00
• 9c81ce4707 [Anthropic API] Preserve image content in tool_result conversion (#19233) Mohammad Miadh Angkad 2026-03-01 04:07:22 +08:00
• a0d8a7ae6d [RadixTree][6/N Refactor]: Refactor SWARadixTree to simplify the computation and alignment of bigram keys. (#19427) zhangheng 2026-02-28 20:01:39 +08:00
• 5705e02d28 Support singleton dimension squeezing in dump comparator (#19566) fzyzcjy 2026-02-28 18:11:46 +08:00
• 80bbd30909 Visualize comparison detailed results in dump comparator (#19565) fzyzcjy 2026-02-28 18:08:16 +08:00
• 40facdb28c Handle recompute and verify closeness in dumper (#19564) fzyzcjy 2026-02-28 18:07:44 +08:00
• 63a4778542 Support non-intrusive arbitrary dumping in dumper and add e2e tests (#19563) fzyzcjy 2026-02-28 18:06:55 +08:00
• ccbc47d6be Update layer id extraction, diffing, empty handling and error sentinel in dump comparator (#19562) fzyzcjy 2026-02-28 18:06:26 +08:00
• 4097eb5ce9 Support patching source code (#19561) fzyzcjy 2026-02-28 18:05:45 +08:00
• b73aa53d7e Enhance metrics in dump comparator (#19560) fzyzcjy 2026-02-28 18:05:19 +08:00
• 706ab9296a Support method decorator for tagging and add minimalistic comparator in dumper (#19559) fzyzcjy 2026-02-28 18:04:54 +08:00
• 9bf3638a25 Support handling arbitrary objects in dump comparator (#19558) fzyzcjy 2026-02-28 18:04:13 +08:00
• b7f13a7b73 [NPU] bugs fix for Deepseek models (#19544) Michelle Wu 2026-02-28 17:26:15 +08:00
• 1eb40d8d45 [NPU]Optimize the PR pipeline to reduce E2E runtime (#18767) Cherry_ming 2026-02-28 17:06:03 +08:00
• 366574b2b8 [PD] Cleanup BootstrapServer init and ready check (#19551) Shangming Cai 2026-02-28 16:41:42 +08:00
• 4ebe9e1e2f [NPU] bugfix: resolve modelslim load weights bug (#19472) Hexq0210 2026-02-28 16:22:45 +08:00
• 53c767d224 [diffusion] Postprocess: implement frame interpolation using RIFE (#19384) Junhao Liu 2026-02-27 22:13:20 -08:00
• b01b07aa16 [diffusion] CI: GT generation flow for diffusion CI (#19236) Yuhao Yang 2026-02-28 14:07:45 +08:00
• b01f3590be [PD] Support PD with context parallel after refactor (#19504) Shangming Cai 2026-02-28 13:11:15 +08:00
• 0404ef9023 [AMD] Add triggering path for multimodal test in pr-test-amd (#19515) YC Tseng 2026-02-28 07:46:37 +03:00
• 79b1d2bac6 [loader] support presharded fused mlp loading (#19519) Yilong Zhao 2026-02-27 20:37:24 -08:00
• 71620122c9 feat(grpc): add multimodal TensorData parsing for vision inference (#19535) Chang Su 2026-02-27 19:29:43 -08:00
• a2ea5941d5 [feat] Support nvfp4 quantized model of Qwen3-Next (#17627) Zheng Duan 2026-02-28 10:28:47 +08:00
• ac400cb7bb [CLI] Add --model-type override and keep launch_server supported (#19523) Liangsheng Yin 2026-02-27 18:16:31 -08:00
• e08ef06758 [Session] Gate streaming sessions with --enable-streaming-session and spec v2 guard (#19531) Liangsheng Yin 2026-02-27 18:14:55 -08:00
• b5a8e4179e [SGL] sync patch: Remove sync points, prefill cudagraph for DP, disable cache reset in mem check (#19190) Leon Gao 2026-02-27 18:11:05 -08:00
• 8c0f2d40bd [CI] Nutanix LoRA adapter removed from HuggingFace (#19533) Yuwei An 2026-02-27 17:46:24 -08:00
• 5f07ff9271 Added the prefill delayer policy: The prefill deplay range is expanded. (#17456) chenxu214 2026-02-28 08:56:49 +08:00
• c6cb0c9649 [Session] Add streaming mode with SessionAwareCache fast path (#19171) Aurick Qiao 2026-02-27 16:31:08 -08:00
• 2bd2b60b5c [CI] Disable test_lora_update: Nutanix LoRA adapter removed from HuggingFace (#19527) Alison Shao 2026-02-27 15:38:19 -08:00
• 9469ad089b Fix nvfp4 weight update (#18085) Ziang Li 2026-02-27 14:55:08 -08:00
• 6ca7da3e7c Fix nightly VLM accuracy: gemma3n TP fixes + removal, latency thresholds (#19401) Alison Shao 2026-02-27 14:24:02 -08:00
• e6da514c2c CI: use 'sglang serve' in CI tests (#18597) yrk111222 2026-02-28 06:00:41 +08:00
• 776709efe8 [3/n] deepseek_v2.py Refactor: Migrate MLA forward method in deepseek_v2.py (#19122) Baizhou Zhang 2026-02-27 13:37:29 -08:00
• 7e46aafebb [AMD] Enable cudagraph for aiter nsa backend and add aiter impl for nsa pr… (#18526) wufann 2026-02-28 05:18:32 +08:00
• 36dc973cbf [HiCache] refactor page_first_direct io kernel (#18113) huangtingwei 2026-02-28 03:43:14 +08:00
• 1b75d0d1a9 Fix BatchMLAPagedAttentionWrapper query/qo_inptr mismatch for EAGLE (#15601) Shu Wang 2026-02-27 13:35:45 -06:00
• 6a1480ce45 Fix HiCacheNixl TypeError: mem_pool_host passed as file_path (#19517) ishandhanani 2026-02-27 12:59:32 -06:00
• 35ef38c61b Remove gpt-oss hybrid swa gate for trtllm_mha (#19079) Mohammad Miadh Angkad 2026-02-28 02:30:00 +08:00
• 1b79934d34 [AMD] Fix AMD CI test of TestToolChoiceLfm2Moe (#19113) Michael 2026-02-27 10:18:15 -08:00
• 2c856c6d27 Allow PR authors to use /rerun-failed-ci on their own PRs (#19496) Alison Shao 2026-02-27 10:14:57 -08:00
• fe4bc8ebd5 [diffusion] fix: MulAdd 4D path (shift indexing) (#18673) R0CKSTAR 2026-02-28 01:52:57 +08:00
• b1249ac909 [Diffusion] [NPU] [CI] fix CI performance (#19486) Makcum888e 2026-02-27 18:23:02 +03:00
• d2885a9094 [Qwen3-Next] Support gdn fused_rms_norm_gated (#19434) Yuan Luo 2026-02-27 23:08:08 +08:00
• ca5f2e2ed1 [diffusion] fix: Support default response_format=url in /v1/images/generations to avoid 400 errors when response_format is omitted (#19360) joesun 2026-02-27 19:47:38 +08:00
• 98e433e305 [PD-Disagg][Fix] Remove 'test_external_dp_routing' from Rust Router constructor parameters. (#19492) Duyi-Wang 2026-02-27 18:40:01 +08:00
• 403195d59d [AMD] [MiniMax-M2.5 Day 0] Add MiniMax-M2.5 nightly accuracy test (#19443) Michael 2026-02-27 02:39:33 -08:00
• f69ca93d49 [AMD] remove redundancy H2D op in aiter attention backend (#19416) AMD-yanfeiwang 2026-02-27 18:15:55 +08:00
• 07da4bed7b [cache] add conservative estimation (#19482) Yilong Zhao 2026-02-27 02:14:46 -08:00
• 9496bbd7b1 [AMD] Use tilelang as default NSA attention backend dispatch on AMD Instinct (#18319) fxmarty-amd 2026-02-27 10:43:34 +01:00
• 9b2fbf7e6a [AMD] Merge Dockerfiles for ROCm (#19203) Alan Kao 2026-02-27 16:53:51 +08:00
• 43fade5f69 [4/N] (Elastic EP) Back up Expert Weights in DRAM (#17374) ympcMark 2026-02-27 15:59:13 +08:00
• eef44ec916 [NPU]kimi k2 thinking bugfix (#19387) KnightLTC 2026-02-27 15:36:38 +08:00
• 4f0f6cd9d0 Add torch.compile support for qwen3-next on CPU (#12444) Cao E 2026-02-27 15:28:03 +08:00
• bc9190435b llama4 npu adapt (#17123) KnightLTC 2026-02-27 14:52:37 +08:00
• f0c2089597 [vlm][internVL] Support processor and embedding inputs for InternVL (#19127) Ken J 2026-02-26 22:46:48 -08:00
• 8a56cc5836 [diffusion] model: Fix a performance bug in the Wan model about usp (#19340) triple-mu 2026-02-27 14:36:23 +08:00
• af3eccc9ab [Perf] Eliminate the slice op for Flashinfer trtllm_fp4_block_scale_moe (#15731) elvischenv 2026-02-27 14:23:40 +08:00
• d566816d83 fix qwen3_vl visual module loading (#19333) wingedge777 2026-02-27 11:44:05 +08:00
• fe971b620a [Diffusion] Add SGL-D diffusion efficient kernel skills (#19473) Xiaoyu Zhang 2026-02-27 11:42:13 +08:00
• 52c8a3632a Fix missing StandardCombineInput import in BF16 flashinfer_trtllm MoE (#19400) Alison Shao 2026-02-26 18:32:08 -08:00
• c2dce06d9f Fix parallel tool call test for speculative decoding variants (#19370) Alison Shao 2026-02-26 18:31:20 -08:00
• 5b5c509480 [NPU][feature adapt]remote load weight feature adp npu (#17968) littleyellowbicycle 2026-02-27 10:10:36 +08:00
• e567215e44 [bugfix]fix fa4 decoding (#19388) Yizhong Cao 2026-02-27 09:48:21 +08:00
• 054bd71086 [sgl-kernel slimming] remove sgl-kernel moe-wna16-marlin (#19379) Xiaoyu Zhang 2026-02-27 09:17:46 +08:00
• 5194eef88b [Fix] A followup fix for TRTLLM BF16 MoE (#15303) Sam (Kesen Li) 2026-02-27 09:14:20 +08:00
• f0d78f2e20 [AMD] Fix weight load shape mismatch for amd dsr1 0528 mxfp4 (#19425) billishyahao 2026-02-27 09:11:59 +08:00
• 4add6ec0f6 Enhance displaying and debuggability in dump comparator (#19466) fzyzcjy 2026-02-27 09:06:19 +08:00
• 23cbbd6d41 Integrate packed data context parallel in dump comparator (#19464) fzyzcjy 2026-02-27 08:28:36 +08:00
• 1a32c0db4d [Minor] Rename misleading chunked to reusing in ReqToTokenPool.alloc() (#19465) Liangsheng Yin 2026-02-26 16:24:03 -08:00
• 8293a914a6 Support token align with packed CP data in dump comparator (#19463) fzyzcjy 2026-02-27 08:12:54 +08:00
• 695e93b91f Make reorderer support packed format with CP in dump comparator (#19462) fzyzcjy 2026-02-27 08:12:18 +08:00
• e3cdf6b1a3 Support CP packed format in unsharder in dump comparator (#19461) fzyzcjy 2026-02-27 08:11:47 +08:00