Commit Graph

Select branches

Hide Pull Requests

cch/megamoe-rc-v0.1

Mono Color

• 67caea6fe4 fix: adding deepseek base tests to b200 (#15915) Douglas Yang 2025-12-26 18:04:25 -08:00
• cd3289c7a4 [BugFix][VLM] Correct weight loading with tie_word_embeddings == False (#15398) Zheng Wengang 2025-12-27 09:30:50 +08:00
• acddb8e0ec Revert embedding integration tests from 5f3a47d8a (#15914) Simo Lin 2025-12-26 20:29:03 -05:00
• 2ec57cefd9 hotfix: add type hints to scheduler mixins (#15916) Cheng Wan 2025-12-26 17:08:11 -08:00
• 988b14ca0e refactor: add type hints to scheduler mixins (#15913) Cheng Wan 2025-12-26 16:50:07 -08:00
• 93495dcac9 Revert "[VLM] Refactor load_mm_data to improve performance" (#15911) Lianmin Zheng 2025-12-26 13:43:18 -08:00
• 886e038329 [model-gateway]: fix crash in embedding worker health check (#15910) Ratish P 2025-12-27 00:57:13 +04:00
• 01bd0d3e8b [Tool Call][DSV32] Streamline function call parameters (#14750) Muqi Li 2025-12-27 03:35:30 +08:00
• 43e1bbc0d5 Revert "[feat] Init support for webui-I2I" (#15906) Lianmin Zheng 2025-12-26 09:53:03 -08:00
• 59b12996fb [diffusion] apps: support I2I tasks in webui (#15778) Li Jinliang 2025-12-26 23:45:11 +08:00
• b70914969b [diffusion] CI: support returning request id from endpoint (#15844) Mick 2025-12-26 23:38:03 +08:00
• 51dbdb2202 [diffusion] improve: improve qwen-image-edit performance to align with LightX2V (#15812) Xiaoyu Zhang 2025-12-26 22:25:10 +08:00
• 8dc6f0fc4d [diffusion] refactor: centralize hardware platform detection and streamline environment variable management (#15842) Mick 2025-12-26 22:16:18 +08:00
• cf34d0ab32 [Fix] assert error in log_prefill_stats (#15881) Huaixin Chang 2025-12-26 19:54:35 +08:00
• 3778c2fc6d [diffusion] CI: fix CI test case skip problem (#15874) Prozac614 2025-12-26 19:42:20 +08:00
• 5d421db883 Tiny log warn users when tracing is automatically disabled (#15889) fzyzcjy 2025-12-26 19:00:37 +08:00
• fe3d47fc9d fix: warn once per env var key (#15846) shuwenn 2025-12-26 18:41:15 +08:00
• ef92b4eb88 [BUGFIX] fix edge case for qwen3-next (#14209) Yi Zhang 2025-12-26 17:14:37 +08:00
• 73c0c66ff8 [diffusion] ZImage support Tensor Parallel (#15849) Annis 2025-12-26 14:58:14 +08:00
• a1e9b4edfa diffusion: support Qwen-Image-Layered (#15817) chhnb 2025-12-26 14:57:01 +08:00
• e75657c839 [Bug] fix piggyback load report return None bug (#15870) Liangsheng Yin 2025-12-26 14:48:45 +08:00
• c28c536c91 Fix swa available memory check (#15867) Ke Bao 2025-12-26 13:19:01 +08:00
• 086813ae8a [VLM] refactor: refactor load_mm_data to improve performance (#14644) Yuan Luo 2025-12-26 13:17:53 +08:00
• 3fd232ad96 [NPU] update Mixed chunk op to FIA (#15518) Hexq0210 2025-12-26 12:17:40 +08:00
• cb1812954a Introduce ModelRunnerKVCacheMixin to simplify the code. (#15821) Liangsheng Yin 2025-12-26 11:17:33 +08:00
• a91e072f33 Use allow auto truncate in the OpenAI API endpoint (#15369) Kangyan-Zhou 2025-12-25 18:45:57 -08:00
• 0271fc3456 Fix prefill num tokens metrics (#15858) fzyzcjy 2025-12-26 10:26:42 +08:00
• 68bece8caa Super tiny move last_prefill_tokens to metrics mixin (#15857) fzyzcjy 2025-12-26 10:26:10 +08:00
• 7b7e357f61 Separate swa and local attention chunk cache eviction (#15820) Ke Bao 2025-12-26 09:34:22 +08:00
• 2f66b0671b Fix chunk_kda_fwd missing argument (#15851) Ke Bao 2025-12-26 09:32:35 +08:00
• 4805118122 ci: add continue-on-error for scheduled PR tests (#15701) Alison Shao 2025-12-25 13:26:14 -08:00
• f2ccc44240 fix: improving format and design (#15791) Douglas Yang 2025-12-25 13:13:27 -08:00
• caa95c7eb4 [model-gateway] update ManualPolicy with header-based routing (#15847) Simo Lin 2025-12-25 16:04:43 -05:00
• 9d878c1f3e Optimize FP8 MLA KV cache writes with Triton kernel (#15522) Hudson Xing 2025-12-25 12:35:39 -08:00
• 8087ef126f Clean up the __init__ of TokenizerManager and DetokenizerManager (#15796) Lianmin Zheng 2025-12-25 10:30:27 -08:00
• 6ef543f99d Use X-SMG-Routing-Key header instead of json body and add tests (#15826) fzyzcjy 2025-12-26 01:58:14 +08:00
• a355911909 [diffusion] improve: improve post-processing by moving compute-intensive tasks to GPU (#15822) Mick 2025-12-26 01:29:04 +08:00
• f3ba711662 fix: change class name of GLM-ASR (#15772) Yuxuan Zhang 2025-12-26 00:15:16 +08:00
• 5c243ba588 Custom All Reduce for Piecewise Cuda Graph (#15356) Yuwei An 2025-12-25 07:55:15 -08:00
• b6702d72cf feat: log request when e2e latency exceeds the specified value (#15759) StonyPort 2025-12-25 23:12:35 +08:00
• c12567274a Tiny change bench-serving to use routing key header (#15827) fzyzcjy 2025-12-25 21:29:49 +08:00
• bb9e6cdf9e [MiMoV2Flash] fix: respect --swa-full-tokens-ratio arg (#15488) Yingchun Lai 2025-12-25 21:02:56 +08:00
• de03b0cd30 [Nemotron 3 Nano] Add triton MoE configs (#15815) roikoren755 2025-12-25 13:31:20 +02:00
• 2a8a785634 [diffusion] log: avoid logging in hot path if unnecessary (#15818) Mick 2025-12-25 18:42:52 +08:00
• f4e835af2f Cleanup ModelRunner (#15802) Liangsheng Yin 2025-12-25 18:13:30 +08:00
• e6ce16a4c2 [diffusion] feat: support TP for Flux.1.dev (#15666) Xiaoyu Zhang 2025-12-25 18:00:44 +08:00
• de2f2880b5 [JIT sgl-kernel] Jit support per tensor quant (#15709) Xiaoyu Zhang 2025-12-25 16:24:37 +08:00
• a89e85e739 [1/N][Sparse With Hicache]: Add Sparse Interface (#14741) zhangheng 2025-12-25 16:12:47 +08:00
• cbf9f13493 Adjust server args for Mimo-v2-flash model (#15803) Ke Bao 2025-12-25 16:04:39 +08:00
• eb3da9c1dd fuse ssm state store into chunk_gated_delta_rule_fwd_h (#15409) Yi Zhang 2025-12-25 16:03:15 +08:00
• 72a980c6d5 ci: migrate MLA tests to test/registered/mla/ (#15798) Alison Shao 2025-12-24 23:29:30 -08:00
• ccf2330b28 Tiny fix missing record_router_upstream_response (#15811) fzyzcjy 2025-12-25 15:20:43 +08:00
• b9af8d2eb9 [VLM] Support apply qk norm in multi cuda streams (#15720) Yuan Luo 2025-12-25 14:35:07 +08:00
• 10a9573efd [diffusion] CI: add tests to validate the size, extension, and format of output images/videos. (#15736) Prozac614 2025-12-25 14:00:22 +08:00
• 49ab72f837 [diffusion] fix: remove unused LoRA application logic from RowParallelLinearWithLoRA class in linear.py (#15801) Prozac614 2025-12-25 13:43:38 +08:00
• 8865424f14 [diffusion] refactor: refactor attention backend checking to use backend enum (#15555) R0CKSTAR 2025-12-25 11:51:40 +08:00
• b311c43d13 Clarify None handling in sglang's environ (#15770) Liangsheng Yin 2025-12-25 11:40:17 +08:00
• 0c39730b18 DP: support piggyback server load report (#11469) Huaixin Chang 2025-12-25 11:35:05 +08:00
• 45adad37d0 Add manual routing policy for router (#15586) fzyzcjy 2025-12-25 11:27:36 +08:00
• 1ba897f330 Tiny refactor select_workers API for future passing more information (#15596) fzyzcjy 2025-12-25 10:31:10 +08:00
• 38dd4fbb66 Add overlap scheduling for embeddings code path (#14032) satyamk7054 2025-12-24 18:24:18 -08:00
• ecd2d09a6c Fix code sync scripts (#15787) Lianmin Zheng 2025-12-24 16:13:23 -08:00
• 92ddc46824 [Auto Sync] Update server_args.py (20251223) (#15700) Lianmin Zheng 2025-12-24 14:55:14 -08:00
• 5454d2a7ae [model-gateway] release smg 0.3.0 (#15781) Simo Lin 2025-12-24 16:23:57 -05:00
• 17b38f88bc Add LoRA metrics for potential auto scaling (#15149) Chenxi Li 2025-12-24 12:13:52 -08:00
• ae434f7821 Move limit-mm-data-per-request to make code clean (#15775) Xinyuan Tong 2025-12-24 18:59:39 +00:00
• 643aeefe6d [model-gateway] Fix logging module name, parse endpoint context, and tokenizer factory (#15782) Simo Lin 2025-12-24 13:54:48 -05:00
• 186a56f6e2 fix(monitoring): update Grafana dashboard metrics prefix from sglang: to sglang_ (#15758) ゆり 2025-12-25 02:43:51 +08:00
• 17e65466de fix: nightly fix b200 gpqa (#15745) Douglas Yang 2025-12-24 10:29:45 -08:00
• 370bd27f3a [model-gateway] Implement Zero-Copy Vision Tensor Access (#15750) Praneth Paruchuri 2025-12-24 22:45:44 +05:30
• b27b5a834e [bug] fix code formatting which blocks ci (#15780) Simo Lin 2025-12-24 12:14:13 -05:00
• 2f7c629235 [model-gateway] Fix IGW routing and optimize RouterManager (#15741) Simo Lin 2025-12-24 11:28:53 -05:00
• 2fb3160598 [NPU] Bug fix in device detect (#14137) Mayyyy 2025-12-25 00:18:39 +08:00
• 8bf7f240b6 [NPU]qwen3 pp bugfix (#15390) Liwansi 2025-12-25 00:14:40 +08:00
• 2c5679f314 [diffusion] refactor: unify the profiling api for all executors (#15718) Mick 2025-12-24 23:26:09 +08:00
• 159b128357 Tiny add flush for CI crash locating (#15769) Liangsheng Yin 2025-12-24 22:47:10 +08:00
• 9338f63f7d Add request counter in addition to existing response counter (#15768) fzyzcjy 2025-12-24 22:22:38 +08:00
• 8196998a03 Tiny add num retracted tokens metric (#15653) fzyzcjy 2025-12-24 22:15:23 +08:00
• aa21c6e3f1 Super tiny code cleanup (#15652) fzyzcjy 2025-12-24 22:15:00 +08:00
• fd4a558e71 Add metrics for having prefill and decode in different ranks (#15752) fzyzcjy 2025-12-24 21:35:35 +08:00
• b3b818fd86 Tiny unify realtime_tokens_total metric (#15747) fzyzcjy 2025-12-24 21:29:43 +08:00
• d5fbbfd988 Retry removing wrong logic about max total token in spec decoding (#15748) Liangsheng Yin 2025-12-24 21:13:26 +08:00
• e245cac0c5 Support JSON format request logging for easier parsing (#15743) fzyzcjy 2025-12-24 19:52:43 +08:00
• c6a6ba4368 Tiny refactor request logger (#15740) fzyzcjy 2025-12-24 19:51:26 +08:00
• d6108166bc Fix smg_http_requests_total semantics (#15655) fzyzcjy 2025-12-24 17:41:01 +08:00
• ddb3970eb4 Tiny env cleanup in deepgemm (#15706) vincentzed 2025-12-24 03:58:46 -05:00
• f65fa04748 [model-gateway]Enable IGW mode with gRPC router and auto enable IGW when service discovery is turned on (#15459) Arthur Cheng 2025-12-24 00:27:58 -08:00
• 7e027691c8 update benchmark README to use --fp8-gemm-backend instead of env var (#15689) Lee Nau 2025-12-24 02:23:31 -05:00
• cb719c74ad [2/N] clean duplicate code of logprob processing in spec. (#15593) Liangsheng Yin 2025-12-24 15:21:13 +08:00
• ff903a7eea Simplify server args (#15704) Lianmin Zheng 2025-12-23 22:46:12 -08:00
• eee3700d84 [diffusion] feat: support lora strength (#15691) Prozac614 2025-12-24 13:31:23 +08:00
• e254cdf326 [CI] Remove pcg-omni-ci (#15656) Yuwei An 2025-12-23 21:28:05 -08:00
• dfb5357448 [diffusion] http-server: relax openai image endpoint's strict content_type limit (#15717) Mick 2025-12-24 13:26:13 +08:00
• 9665574937 [docs] major SGL Model Gateway documentation update (#15715) Simo Lin 2025-12-23 20:26:09 -08:00
• ac320a6f04 Move some quant args to its own section in environ variables doc (#15722) vincentzed 2025-12-23 23:11:08 -05:00
• 6292c24437 Tiny fix test eagle infer b. (#15716) Liangsheng Yin 2025-12-24 11:57:15 +08:00
• 5f5a567768 Tiny add flush in the suite partition status print. (#15719) Liangsheng Yin 2025-12-24 11:55:59 +08:00
• 3bf07c684f [Feature][MM] split the images of one request into multiparts (#11828) Xuchun Shang 2025-12-24 11:22:04 +08:00
• e7b09efc0a [AMD] Add AMD Nightly Performance & VLMs Accuracy Tests (#15500) michael-amd 2025-12-23 19:03:27 -08:00
• 99d3bcdfed [model-gateway] add back router worker health metric and fix init state (#15622) fzyzcjy 2025-12-24 10:50:38 +08:00