Skip to content

Pull requests: SemiAnalysisAI/InferenceX

New pull request New
105 Open 2,058 Closed
105 Open 2,058 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Reviews
Filter by reviews
No reviews Review required Approved review Changes requested
Assignee
Filter by who’s assigned
Assigned to nobody Loading
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Pull requests list

perf: add GB200 DeepSeek V4 FP4 vLLM MTP AgentX agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2567 opened Aug 11, 2026 by cquil11 Collaborator Loading…
4
perf: add GB200 Qwen3.5 FP4 SGLang MTP AgentX agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2568 opened Aug 11, 2026 by cquil11 Collaborator Loading…
6
[AMD] [AgentX] Add MI325X Qwen3.5 FP8 SGLang MTP / [AMD] [AgentX] 新增 MI325X Qwen3.5 FP8 SGLang MTP agentx AgentX benchmarks, recipes, and infrastructure agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable AMD full-sweep-enabled
#2566 opened Aug 11, 2026 by cquil11 Collaborator Loading…
2
MiniMax-M3 H200 AgentX EAGLE3 tuning on vLLM v0.27.1 agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2565 opened Aug 11, 2026 by cquil11 Collaborator Loading…
5
MiniMax-M3 H100 AgentX EAGLE3 tuning on vLLM v0.27.1 agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2564 opened Aug 11, 2026 by cquil11 Collaborator Loading…
5
[AMD] [AgentX] Add MI300X Qwen3.5 FP8 SGLang MTP / [AMD] [AgentX] 新增 MI300X Qwen3.5 FP8 SGLang MTP agentx AgentX benchmarks, recipes, and infrastructure agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable AMD full-sweep-enabled
#2563 opened Aug 11, 2026 by cquil11 Collaborator Loading…
4
[AMD] [AgentX] Add MI355X Qwen3.5 FP4 SGLang MTP / [AMD] [AgentX] 新增 MI355X Qwen3.5 FP4 SGLang MTP agentx AgentX benchmarks, recipes, and infrastructure AMD
#2562 opened Aug 11, 2026 by cquil11 Collaborator Loading…
5
chore(agentx): sync support with production results
#2561 opened Aug 11, 2026 by cquil11 Collaborator Loading…
6
Add DSV4 B200 disaggregated Dynamo SGLang STP configuration / 新增 DSV4 B200 分离式 Dynamo SGLang STP 配置 full-sweep-enabled
#2560 opened Aug 11, 2026 by RohitNagraj Collaborator Loading…
4 of 9 tasks
3
feat: validate ATOM and Infera on MI300X and MI355X
#2558 opened Aug 11, 2026 by cquil11 Collaborator Loading…
7
[AMD] Qwen3.5-FP8 MI355X SGLang disagg perf tuning: image bump to v0.5.16, TP4P+TP8D baseline, add MTP / Qwen3.5-FP8 MI355X SGLang disagg 性能调优：镜像升级至v0.5.16，TP4P+TP8D基线优化，新增MTP配置 full-sweep-fail-fast
#2557 opened Aug 11, 2026 by ChangLiu0709 Collaborator Loading…
9
Runner reconciliation
#2556 opened Aug 11, 2026 by cquil11 Collaborator • Draft
[AMD][AgentX] Port MI355X DeepSeek V4 MTP disagg to srt-slurm full-sweep-enabled
#2555 opened Aug 11, 2026 by cquil11 Collaborator Loading…
8
Add DSV4 B200 disaggregated Dynamo SGLang MTP configuration / 新增 DSV4 B200 分离式 Dynamo SGLang MTP 配置 full-sweep-enabled
#2554 opened Aug 11, 2026 by RohitNagraj Collaborator Loading…
4 of 9 tasks
6
Add DeepSeek-V4-Pro FP4 B200 Dynamo TensorRT-LLM recipes / 添加 DeepSeek-V4-Pro FP4 B200 Dynamo TensorRT-LLM 配置 full-sweep-enabled
#2552 opened Aug 10, 2026 by RohitNagraj Collaborator Loading…
7
[AMD] Add kimik3-fp4-mi355x-vllm-agentic-dspark (AgentX, conc 1/2/4/8) full-sweep-fail-fast
#2551 opened Aug 10, 2026 by Fangzhou-Ai Collaborator • Draft
3 tasks
1
perf(qwen3.5-fp4-b200-sglang-mtp): add TEP2 arm full-sweep-enabled
#2550 opened Aug 10, 2026 by hshrivastava-droid Collaborator Loading…
9
Validate vLLM Router on GB200: DEP4, DEP8, and 1P/2D agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2549 opened Aug 10, 2026 by cquil11 Collaborator Loading…
16
Add pinned Kimi Vendor Verifier smoke / 添加固定版本的 Kimi Vendor Verifier 冒烟评估 full-sweep-fail-fast
#2547 opened Aug 10, 2026 by adibarra Collaborator Loading…
11
Aggregate E2E normalized interactivity in AgentX CI results / 在 AgentX CI 结果中聚合端到端归一化交互性
#2544 opened Aug 10, 2026 by Duyi-Wang Collaborator Loading…
1
Validate official vLLM Router with srt-slurm agentx-fast Run AgentX throughput with 1 warmup request per lane and a 20-minute profile; not reusable full-sweep-enabled
#2543 opened Aug 10, 2026 by cquil11 Collaborator Loading…
16
feat(amd): orchestrate AMD clusters with srt-slurm
#2542 opened Aug 10, 2026 by cquil11 Collaborator Loading…
17 tasks done
6
perf(agentx): refresh DeepSeek V4 MTP frontier on MI355X full-sweep-enabled
#2540 opened Aug 9, 2026 by cquil11 Collaborator Loading…
7 of 9 tasks
27
perf(agentx): add DeepSeek V4 FP8 MTP frontier on MI325X full-sweep-enabled
#2541 opened Aug 9, 2026 by cquil11 Collaborator Loading…
5 of 8 tasks
29
fix(evals): preserve GPQA reasoning responses / 修复 GPQA 推理响应截断
#2537 opened Aug 9, 2026 by JordanNanos Collaborator Loading…
ProTip! What’s not been updated in a month: updated:<2026-07-11.