GitHub / SemiAnalysisAI/InferenceX / commits
| SHA | Message | Author | Date | Stats |
|---|---|---|---|---|
| ca4d1d00 |
Update DeepSeek-V4-Pro (DSV4) FP4 GB300 Dynamo-SGLang AgentX MTP configs / 更新...
Co-authored-by: weireweire <2****e@u****m>, Rohit Pujar Nagraj <r****j@n****m>, Rohit Nagraj <r****9@g****m>, Ankur-singh <a****h@n****m>, Cameron Quilici <c****i@g****m> |
weireweire <w****l@n****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 5984ac8c |
docs: deprecate dsv4 Single-turn 8k1k on 2026-09-08, keep AgentX [skip-sweep]...
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m> |
functionstackx <4****x@u****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| c1f5be9b |
perf(gb300): Add more AgentX vLLM MTP aggregate and P/D recipes / 增加更多 GB300 ...
Co-authored-by: Cameron Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 2251a611 | [Power] Require GB multinode telemetry / 强制启用 GB 多节点功耗采集 (#2687) |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 26265c79 |
[AgentX] DeepSeek-V4 B300 SGLang update (#2701)
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Cameron Quilici <c****i@g****m> |
Yuwei An <a****9@g****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 455523ef | [CI] Repair stale runner git state before checkout / checkout 前自愈 runner 陈旧 g... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 76ff8a6f | [Power] Require GLM-5.2 and add DSV4 H200 telemetry / 强制 GLM-5.2 并新增 DSV4 H20... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 12e4bd8a | [Power] Wire H200 multinode AgentX power / 接入 H200 多节点 AgentX 功耗链路 (#2683) |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 29fe352b |
Refresh the DSv4 MI355X SGLang AgentX arm to v0.5.18 and split TP4/TP8 by con...
Co-authored-by: Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Thomas Wang <1****h@g****m> |
karverma-amd <k****a@a****m>
Committed by: GitHub <n****y@g****m> |
4 days ago | |
| 0b0138fd | [TileRT][AgentX] Add glm5.1-fp8-b200-tilert-agentic (#2650) |
Bryan Shan <5****r@u****m>
Committed by: GitHub <n****y@g****m> |
5 days ago | |
| 843c6254 |
Serve the dsv4 B200 TP8 agentic arm with FULL_AND_PIECEWISE cudagraphs (#2707)
Co-authored-by: Ankur-singh <a****h@n****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
7 days ago | |
| 1845f605 | Refresh B300 MiniMax M3 TRT AgentX with stream interval 20 (#2685) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
7 days ago | |
| 49ac5c0a |
Add MiniMax-M3 NVFP4 GB300 vLLM Disaggregated AgentX with EAGLE3-GQA MTP / 新增...
Co-authored-by: Oseltamivir <5****r@u****m>, Ankur-singh <a****h@n****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
7 days ago | |
| e0e8bfdd | fix(config): mark GLM-5.2 TRT AgentX host KV offload (#2709) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| e11dd394 | fix(config): mark DSV4 TRT AgentX host KV offload (#2708) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| bd171877 |
feat(config): add GLM-5.2 GB300 Dynamo-TRT AgentX / 添加 GLM-5.2 GB300 Dynamo-T...
Co-authored-by: Cameron Quilici <6****1@u****m>, Claude Fable 5 <n****y@a****m>, Ankur-singh <a****h@n****m>, Cameron Quilici <c****i@g****m>, claude[bot] <4****]@u****m>, Claude Fable 5 <n****y@a****m> |
Rohit Nagraj <r****9@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| fbb480e5 | [AgentX] DeepSeek-V4 B200 SGLang update (#2703) |
Po-Han Huang (NVIDIA) <5****h@u****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| c0c91567 | fix: revert required srt frontend variables (#2706) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| de871da9 |
perf(gb300): add AgentX aggregate TP4 points / 添加 GB300 AgentX 聚合式 TP4 数据点 (#...
Co-authored-by: Cameron Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| 43996b8d |
perf(agentx): Add B300 vLLM MTP TP4 data points (#2647)
Co-authored-by: Cameron Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| 6816ed05 | config(dsv4): add GB300 Dynamo-TRT AgentX points / 添加 GB300 Dynamo-TRT AgentX... |
Rohit Nagraj <r****9@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| 056876ec | Kimi K3 vllm GB200 Agentx update (#2700) |
Hanjie Qiu <5****q@u****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| e944302e |
AMD AgentX MI355X: GPU sanity dedup, HiCache tuning, node-exclude scoping, an...
Co-authored-by: Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Chun Fang <c****g@a****m>, billishyahao <b****e@a****m>, Cameron Quilici <c****i@g****m> Signed-off-by: Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m> |
Theresa Shan <t****n@a****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| 38fbc472 |
feat(agentx): add Kimi-K3 FP4 MI355X ATOM DSpark recipe (#2697)
Co-authored-by: Claude Opus 4.6 <n****y@a****m>, seungrokj <s****g@a****m>, Claude Opus 4.6 <n****y@a****m>, seungrokj <1****j@u****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
8 days ago | |
| c2b58030 |
[AMD][AgentX] Add DeepSeek-V4-Pro FP4 MI355X ATOM MTP (#2698)
Co-authored-by: Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m> |
seungrokj <1****j@u****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| bda9a7a4 |
[Klaud Cold] docs: add AgentX, GLM5.3, and Kimi K3 2.8T to README news / 在 RE...
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m> |
functionstackx <4****x@u****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| cbfb4a9d | Fix agentic eval for KV-offload entries (pass backend name, not object) (#2692) |
seungrokj <1****j@u****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| 108bf4f2 |
Refresh B200 MiniMax M3 TRT AgentX with stream interval 20 (#2686)
Co-authored-by: functionstackx <4****x@u****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| d658eac9 |
[AMD][MI35X] Bump Qwen3.5 MXFP4 MI355X SGLang AgentX image to v0.5.17-2026081...
Co-authored-by: functionstackx <4****x@u****m> |
jacky.cheng <y****e@a****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| dedca9eb | Fix agentic eval speculative decoding / 修复 AgentX 评测的推测解码传递 (#2689) |
Bryan Shan <5****r@u****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| 64c7b612 |
Kimi K3 vllm GB200 Agentx (#2639)
Co-authored-by: OpenAI Codex <c****x@o****m>, OpenAI Codex <c****x@o****m>, OpenAI Codex <c****x@o****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m> Signed-off-by: Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m> |
Wei Zhao <5****8@u****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| 2719ad8c |
Refresh GB200 Qwen3.5 AgentX Pareto / 刷新 GB200 Qwen3.5 AgentX Pareto 配置 (#2660)
Co-authored-by: functionstackx <4****x@u****m> |
Ankur Singh <a****h@n****m>
Committed by: GitHub <n****y@g****m> |
9 days ago | |
| 5eb64485 |
feat(config): add GLM-5.2 FP4 GB200 SGLang AgentX MTP / 添加 GLM-5.2 FP4 GB200 ...
Co-authored-by: functionstackx <4****x@u****m> |
Rohit Nagraj <r****9@g****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| bf317e63 |
Add Kimi K3 FP4 B200 PP2 agentic sweep (#2672)
Co-authored-by: functionstackx <4****x@u****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| f0711382 |
docs(golden_al): document ATOM's forced acceptance-length knob / 补充 ATOM 的强制接...
Co-authored-by: Cursor <c****t@c****m> |
Hexiang Wang <5****u@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| def9a650 | [Power] feat: extend dcgm energy lanes to gb dsv4/qwen3.5 fp4 / 扩展 dcgm 能耗采集到... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| c5f3f82b | [Power] feat: measure GPU power for single-node AgentX runs / 为单机 AgentX 运行接入... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| acf3b878 | feat(power): stamp whole-deployment semantics with a schema version (#2599) |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 1b332e4c |
[AMD] [AGENTX] GLM-5.2 FP4 MI355X ATOM Agentic MTP (#2576)
Co-authored-by: Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, zhuyuhua-v <y****u@a****m>, adibarra <9****a@u****m> Signed-off-by: zhuyuhua-v <y****u@a****m>, zhuyuhua-v <y****u@a****m>, zhuyuhua-v <y****u@a****m> |
seungrokj <1****j@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 81484bdd |
perf(dsv4): update B200 SGLang AgentX HiCache MTP DEP8 configuration / 更新 B20...
Co-authored-by: functionstackx <4****x@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 762ae70e |
Add MiniMax-M3 NVFP4 B300 TensorRT-LLM AgentX with EAGLE3-GQA MTP / 新增 MiniMa...
Co-authored-by: functionstackx <4****x@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 9c3f6697 |
[AMD][AgentX] Update DSv4 MI355X vLLM MTP flags on the 2026-08-17 ROCm nightl...
Co-authored-by: Claude Opus 5 <n****y@a****m>, Jiahui Cao <j****o@c****l>, Claude Opus 5 <n****y@a****m>, Fangzhou Ai <3****i@u****m>, Chun Fang <c****g@a****m>, functionstackx <4****x@u****m> |
jiacao-amd <j****o@a****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| a030e035 |
Add MiniMax-M3 NVFP4 B200 TensorRT-LLM AgentX with EAGLE3-GQA MTP / 新增 MiniMa...
Co-authored-by: functionstackx <4****x@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 2ddf1043 |
[AMD] Qwen3.5-FP8 MI355X SGLang disagg baseline: image bump to v0.5.16, TP4P+...
Co-authored-by: Cursor <c****t@c****m>, Cursor <c****t@c****m>, Chun Fang <c****g@a****m>, functionstackx <4****x@u****m> |
ChangLiu0709 <c****4@a****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 6fd12b47 |
Kimi K3 vllm GB300 Agentx (#2637)
Co-authored-by: OpenAI Codex <c****x@o****m>, OpenAI Codex <c****x@o****m>, OpenAI Codex <c****x@o****m>, OpenAI Codex <c****x@o****m>, Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, functionstackx <4****x@u****m> Signed-off-by: Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m>, Wei Zhao <5****8@u****m> |
Wei Zhao <5****8@u****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 16f273c8 | fix(agentx): drop hung Qwen3.5 concurrency points [skip-sweep] (#2661) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| bb91726d |
perf: tune GLM-5.2 B200 c12/c16 HiCache / 性能：调优 GLM-5.2 B200 c12/c16 HiCache ...
Co-authored-by: functionstackx <4****x@u****m> |
Ankur Singh <a****h@n****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| 13162e8d |
perf: HiCache tuning for glm5.2-fp4-b300-sglang-agentic-mtp / 性能：glm5.2-fp4-b...
Co-authored-by: functionstackx <4****x@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
10 days ago | |
| b9c9f925 |
Add acceptance length reference values for deepseek model (#2626)
Co-authored-by: Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, functionstackx <4****x@u****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
11 days ago | |
| 4b719316 |
Add Qwen3.5 H200 MTP config / 新增 Qwen3.5 H200 MTP 配置 (#2592)
Co-authored-by: Kedar Potdar <k****r@n****m>, adibarra <9****a@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
11 days ago | |
| 9ddcc1fe |
fix: recover PR 2619 production ingest / 修复 PR 2619 生产环境写入 (#2636)
Co-authored-by: Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Xin Li <x****i@n****m>, Claude Sonnet 5 <n****y@a****m> |
Alec Ibarra <9****a@u****m>
Committed by: GitHub <n****y@g****m> |
11 days ago | |
| 5caeb6be |
[AgentX] Mirror GB300 MTP tuning (#2571) to GB200 with 4-GPU->8-GPU workers (...
Co-authored-by: Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Claude Sonnet 5 <n****y@a****m>, Cameron Quilici <c****i@g****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
11 days ago | |
| a7690a81 |
perf(agentx): retune B200 vLLM MTP aggregate (#2621)
Co-authored-by: Cameron Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
12 days ago | |
| 88bd4828 | fix(gb200): skip srtctl preflight on all paths (#2615) |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
12 days ago | |
| 3d558156 | [NV][AgentX] Add GB200 GLM-5.2 FP4 MTP sweep (#2620) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
12 days ago | |
| 384d0c44 |
Add Qwen3.5-397B-A17B-NVFP4-V2 GB300 TRT-LLM disaggregated AgentX (#2612)
Co-authored-by: Xin Li <x****i@n****m>, Ankur-singh <a****h@n****m>, Cameron Quilici <c****i@g****m> |
Richard Huo <r****o@n****m>
Committed by: GitHub <n****y@g****m> |
13 days ago | |
| cb8a603a |
[NV][AgentX] Add Kimi-K3 B300 vLLM DSpark sweep / [NV][AgentX] 新增 Kimi-K3 B30...
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, Claude Opus 5 (1M context) <n****y@a****m>, Cameron Quilici <c****i@g****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
13 days ago | |
| f8dff19a | Add GLM-5.2-NVFP4 GB300 SGLang AgentX recipes / 新增 GLM-5.2-NVFP4 GB300 SGLang... |
Sahithi Chigurupati <5****i@u****m>
Committed by: GitHub <n****y@g****m> |
13 days ago | |
| 645d519c | feat(minimaxm3): refresh B200 AgentX Pareto points with EAGLE3-GQA / 使用 EAGLE... |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
13 days ago | |
| 4a5cc44b |
feat(minimaxm3): refresh B300 AgentX Pareto points with EAGLE3-GQA / 使用 EAGLE...
Co-authored-by: Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, Bryan Shan <5****r@u****m> |
Xin Li <1****w@u****m>
Committed by: GitHub <n****y@g****m> |
13 days ago | |
| 5b9af262 |
Add GB200 MiniMax-M3 FP4 AgentX sweep (#2609)
Co-authored-by: functionstackx <4****x@u****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
14 days ago | |
| 35d731a9 | feat(ci): support additive append-only curve updates (#2613) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
14 days ago | |
| edcc6032 | chore(agentx): sync support with production results (#2561) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
14 days ago | |
| d136fa6b |
[AgentX] Tune DSV4 GB300 MTP / 调优 DSV4 GB300 MTP (#2571)
Co-authored-by: Cam Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| 731d40ce | [NV] Add SpeedBench AL collectors for DSV4-Pro-0813 on B300 (#2526) |
Albert Cheng <a****g@n****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| 1bb270da | perf: add GB200 DeepSeek V4 FP4 vLLM MTP AgentX (#2567) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| 489bcbb6 | feat(agentx): aggregate E2E normalized interactivity (#2544) |
Duyi-Wang <d****g@a****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| a390b529 |
[AMD] Add MI355X DeepSeek-V4-Pro Sglang Single Node AgentX recipes (#2600)
Co-authored-by: seungrokj <1****j@u****m> |
billishyahao <b****e@a****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| d4f825b1 | Revert "[CI] Add weekly overview snapshot sweep / [CI] 新增每周 overview snapshot... |
Bryan Shan <5****r@u****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| ff4a312f | [CI] Add weekly overview snapshot sweep / [CI] 新增每周 overview snapshot 定时 swee... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
15 days ago | |
| cd9f3d56 | [AMD][AgentX] Add MI300X MiniMax-M3 EAGLE3 MXFP8 (#2580) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
16 days ago | |
| 88500fbb | [AMD] [AgentX] Add MiniMax-M3 MXFP8 EAGLE3 on MI325X (#2579) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
16 days ago | |
| c9544dbd | Refresh DeepSeek-V4 B200 SGLang AgentX MTP (#2578) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
16 days ago | |
| 904433f1 | Refresh DeepSeek-V4 B300 SGLang AgentX MTP (#2577) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
16 days ago | |
| 0a826be3 |
perf(qwen3.5-fp4-b200-sglang-mtp): add TEP2 arm on NVFP4-V2 / perf(qwen3.5-fp...
Co-authored-by: Ankur-singh <a****h@n****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
16 days ago | |
| a007419c |
[AMD] [AGENTX] GLM-5.2 MTP5 tune: raise acc-len and drop hicache at low conc ...
Co-authored-by: Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, Claude Sonnet 4.6 <n****y@a****m>, ajith-sirra-amd <1****d@u****m>, functionstackx <4****x@u****m> |
giovanniguastiamd <g****i@a****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 1446647b |
Add DSV4 B200 disaggregated Dynamo SGLang MTP configuration / 新增 DSV4 B200 分离...
Co-authored-by: adibarra <9****a@u****m> |
Rohit Nagraj <r****9@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 39324e44 |
[AMD][AgentX] Add Kimi-K3 MXFP4 MI355X vLLM agentic MTP recipe (#2508)
Co-authored-by: Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, Cameron Quilici <c****i@g****m> |
seungrokj <1****j@u****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| a6e5d2b1 | [AMD] [AgentX] Add MI355X Qwen3.5 FP4 SGLang MTP (#2562) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 80cce67c |
MiniMax-M3 H200 AgentX EAGLE3 tuning on vLLM v0.27.1 (#2565)
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, functionstackx <4****x@u****m>, Claude Opus 5 (1M context) <n****y@a****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 4506c842 | [AMD] [AgentX] Add MI325X Qwen3.5 FP8 SGLang MTP (#2566) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 6a9c6238 | perf: add GB200 Qwen3.5 FP4 SGLang MTP AgentX (#2568) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 149a11d7 |
MiniMax-M3 H100 AgentX EAGLE3 tuning on vLLM v0.27.1 (#2564)
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, functionstackx <4****x@u****m>, Claude Opus 5 (1M context) <n****y@a****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| db584b40 | [AMD] [AgentX] Add MI300X Qwen3.5 FP8 SGLang MTP (#2563) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 73bd8e20 |
[AMD] [AGENTX] MiniMax-M3 FP4 MTP on MI355X (#2487)
Co-authored-by: Claude Opus 5 (1M context) <n****y@a****m>, Sirra <a****a@a****m>, functionstackx <4****x@u****m>, Claude Opus 5 (1M context) <n****y@a****m>, Cam Quilici <c****i@g****m> Signed-off-by: Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Sirra <a****a@a****m>, Cam Quilici <c****i@g****m> |
ajith-sirra-amd <1****d@u****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 363bb3da | ci: make staging comments English-only (#2573) |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| fe433b19 |
fix(agentx): bump AIPerf and propagate MI355X failures (#2572)
Signed-off-by: Cam Quilici <c****i@g****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 691b8f79 |
[AgentX] Tune Kimi K3 DSpark on B200 (#2475)
Co-authored-by: adibarra <9****a@u****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 2346b5c2 |
[NV] Add H200 DeepSeek-V4-Pro AgentX recipes / [NV] 添加 H200 DeepSeek-V4-Pro A...
Co-authored-by: Ankur-singh <a****h@n****m>, Cameron Quilici <c****i@g****m> |
Sahithi Chigurupati <5****i@u****m>
Committed by: GitHub <n****y@g****m> |
17 days ago | |
| 28f934ce |
[AgentX] Refresh Kimi K3 DSpark on B300 / [AgentX] 刷新 B300 上的 Kimi K3 DSpark ...
Co-authored-by: adibarra <9****a@u****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| a5ec626d |
[AMD] [AGENTX] GLM-5.2 FP4 MI355X SGLang Agentic MTP (#2488)
Co-authored-by: Claude Sonnet 4.6 <n****y@a****m>, Claude Opus 4.6 <n****y@a****m>, ajith-sirra-amd <1****d@u****m>, Claude Sonnet 4.6 <n****y@a****m>, seungrokj <1****j@u****m>, seungrokj <s****g@a****m> Signed-off-by: Giovanni Guasti <g****i@a****m>, Giovanni Guasti <g****i@a****m>, Giovanni Guasti <g****i@a****m>, Giovanni Guasti <g****i@a****m> |
giovanniguastiamd <g****i@a****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| c62f6e33 |
Retune DSV4 B300 AgentX MTP sweep / 调优 DSV4 B300 AgentX MTP 扫描 (#2536)
Co-authored-by: Cameron Quilici <c****i@g****m> |
Yifan Qiao <y****o@i****i>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| 4ef13688 | Emit per-role average power on multinode disaggregated runs / 多节点分离式运行输出每角色平均... |
Wenyao Gao <1****8@u****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| 23f879df |
[AMD][AgentX] DSv4 MI355X agentic disagg: add EAGLE/MTP speculative-decoding ...
Co-authored-by: Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cam Quilici <c****i@g****m> Signed-off-by: Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m>, Theresa Shan <t****n@a****m> |
Theresa Shan <t****n@a****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| 7df19b5b |
Qwen3.5 FP4 GB300 disaggregated Dynamo-SGLang: update SGLang to v0.5.14-cu130...
Co-authored-by: Ankur-singh <a****h@n****m>, adibarra <9****a@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| 4830e334 |
[NV] Add GB300 Qwen3.5 NVFP4 Dynamo-TensorRT-LLM disaggregated benchmarks / [...
Co-authored-by: Ankur-singh <a****h@n****m>, adibarra <9****a@u****m> |
Richard Huo <r****o@n****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| f4eaa398 |
Switch Qwen3.5 FP4 SGLang configs to NVFP4 V2 / 将 Qwen3.5 FP4 SGLang 配置切换到 NV...
Co-authored-by: Claude Opus 4.7 <n****y@a****m>, Claude Opus 4.7 <n****y@a****m>, Ankur-singh <a****h@n****m>, adibarra <9****a@u****m> |
hshrivastava-droid <h****a@n****m>
Committed by: GitHub <n****y@g****m> |
18 days ago | |
| f064a86f |
[NV] GB300 DeepSeek-V4-Pro AgentX Dynamo-SGLang session-routing upgrade / [NV...
Co-authored-by: Ankur-singh <a****h@n****m>, weireweire <2****e@u****m>, Xin Li <x****i@n****m>, Xin Li <1****w@u****m> |
Sahithi Chigurupati <5****i@u****m>
Committed by: GitHub <n****y@g****m> |
19 days ago | |
| 12ed4133 |
[NV] llm-d-vllm: optimize DSv4-Pro GB200 recipe configs (#2498)
Co-authored-by: Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cursor <c****t@c****m>, Cameron Quilici <c****i@g****m>, Cursor <c****t@c****m> |
Ilya Markov <m****7@g****m>
Committed by: GitHub <n****y@g****m> |
19 days ago | |
| b7e8b0e3 |
chore(agentx): bump AIPerf to agentx-v1.0.4 (#2548)
Signed-off-by: Cam Quilici <c****i@g****m> |
Cameron Quilici <c****i@g****m>
Committed by: GitHub <n****y@g****m> |
19 days ago |