Skip to content

Pull requests: SemiAnalysisAI/InferenceX

New pull request New
198 Open 2,704 Closed
198 Open 2,704 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

Add H100 TensorRT-LLM ModelScope coverage / 添加 H100 TensorRT-LLM ModelScope 覆盖 full-sweep-fail-fast
#3323 opened Sep 20, 2026 by functionstackx Collaborator Loading…
8
[Klaud Cold] Add GB300 TP2 DeepSeek-V4.1-Flash vLLM AgentX arm with Engram host offload / 新增 Engram 主机卸载的 GB300 TP2 DeepSeek-V4.1-Flash vLLM AgentX 臂 full-sweep-enabled
#3321 opened Sep 20, 2026 by functionstackx Collaborator Loading…
6 tasks
3
[Klaud Cold] Add GB200 TP2 DeepSeek-V4.1-Flash vLLM AgentX arm with Engram host offload / 新增 Engram 主机卸载的 GB200 TP2 DeepSeek-V4.1-Flash vLLM AgentX 臂 full-sweep-enabled
#3320 opened Sep 20, 2026 by functionstackx Collaborator Loading…
6 tasks
2
[Klaud Cold] Update dsr1-fp8-mi325x-sglang SGLang image to v0.5.20-rocm720-mi30x / 将 dsr1-fp8-mi325x-sglang 的 SGLang 镜像更新至 v0.5.20-rocm720-mi30x
#3318 opened Sep 20, 2026 by adibarra Collaborator • Draft
2
[Klaud Cold] Update dsr1-fp8-b200-trt TensorRT-LLM image to 1.3.0rc27 / 将 dsr1-fp8-b200-trt 的 TensorRT-LLM 镜像更新至 1.3.0rc27
#3317 opened Sep 20, 2026 by adibarra Collaborator • Draft
2
[PowerX] skip unreadable in-window power cells instead of voiding the point / 跳过窗口内读不出的功耗单元格而非作废整个点 full-sweep-enabled priority Preempt other runs on this sweep's runners; restore them at the end (org members only) skip_queue
#3316 opened Sep 20, 2026 by edwingao28 Collaborator Loading…
6 of 10 tasks
3
[AgentX GB300] refresh the full Kimi-K3 curve so every point carries measured power / 重测完整曲线使每个点都带实测功耗 full-sweep-enabled priority Preempt other runs on this sweep's runners; restore them at the end (org members only) skip_queue
#3315 opened Sep 20, 2026 by edwingao28 Collaborator • Draft
1 of 5 tasks
2
[Klaud Cold] Update qwen3.5-fp4-mi355x-sglang SGLang image to v0.5.20-rocm720-mi35x-20260919 / 将 qwen3.5-fp4-mi355x-sglang 的 SGLang 镜像更新至 v0.5.20-rocm720-mi35x-20260919
#3311 opened Sep 20, 2026 by adibarra Collaborator • Draft
2
[Klaud Cold] Add H200 TP4 DeepSeek-V4.1-Flash vLLM AgentX arm with Engram host offload / 新增 Engram 主机卸载的 H200 TP4 DeepSeek-V4.1-Flash vLLM AgentX 臂 full-sweep-enabled
#3310 opened Sep 20, 2026 by functionstackx Collaborator Loading…
6 tasks
5
[AgentX H200] refresh the full Kimi-K3 curve so every point carries measured power / 重测完整曲线使每个点都带实测功耗 full-sweep-enabled priority Preempt other runs on this sweep's runners; restore them at the end (org members only) skip_queue
#3309 opened Sep 20, 2026 by edwingao28 Collaborator • Draft
1 of 5 tasks
2
[Klaud Cold] MI325X Qwen3.8-27B BF16 MTP, synthetic AL 2.51 / MI325X Qwen3.8-27B BF16 MTP，合成 AL 2.51 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3308 opened Sep 20, 2026 by Oseltamivir Collaborator • Draft
2
[Klaud Cold] MI300X Qwen3.8-27B BF16 MTP 1k1k/8k1k, AL 2.51 / MI300X Qwen3.8-27B BF16 MTP 1k1k/8k1k，AL 2.51 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3307 opened Sep 20, 2026 by Oseltamivir Collaborator • Draft
2
[Klaud Cold] H200 Qwen3.8-27B BF16 MTP, synthetic AL 2.51 / H200 Qwen3.8-27B BF16 MTP，合成 AL 2.51 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3306 opened Sep 20, 2026 by Oseltamivir Collaborator • Draft
2
[Klaud Cold] H100 Qwen3.8-27B BF16 MTP 1k1k/8k1k, AL 2.51 / H100 Qwen3.8-27B BF16 MTP 1k1k/8k1k，AL 2.51 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3305 opened Sep 20, 2026 by Oseltamivir Collaborator • Draft
2
Add Qwen3.8-27B native MTP golden AL/AR for FP8 and BF16 / 新增 FP8 和 BF16 的 Qwen3.8-27B 原生 MTP 黄金 AL/AR
#3304 opened Sep 20, 2026 by Oseltamivir Collaborator • Draft
1
[Klaud Cold] Document lessons from the September 2026 DSpark and GLM-5.2 AgentX sweeps / 记录 2026 年 9 月 DSpark 与 GLM-5.2 AgentX sweep 的经验教训
#3302 opened Sep 20, 2026 by functionstackx Collaborator Loading…
[wip ai slop] Implement Phase 1 native H100 pilot / 实现阶段 1 原生 H100 试点 full-sweep-enabled
#3299 opened Sep 19, 2026 by functionstackx Collaborator Loading…
9
[wip ai slop] Add trusted Phase 1 receipt controls / 新增阶段 1 可信回执控制
#3298 opened Sep 19, 2026 by functionstackx Collaborator Loading…
2
perf(b300): update vLLM AgentX to DSpark6 / 更新 B300 vLLM AgentX 至 DSpark6
#3297 opened Sep 19, 2026 by ivanium Collaborator Loading…
5 of 6 tasks
5
[PowerX] measure Grace CPU-side power on NVL72 power lanes / 在 NVL72 功耗 lane 上实测 Grace CPU 侧功耗 full-sweep-fail-fast
#3296 opened Sep 19, 2026 by edwingao28 Collaborator • Draft
7 of 10 tasks
1
[Klaud Cold] Add MI325X Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k / 新增 MI325X Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k 配方 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3295 opened Sep 19, 2026 by functionstackx Collaborator Loading…
5
[Klaud Cold] Add MI300X Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k / 新增 MI300X Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k 配方 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3294 opened Sep 19, 2026 by functionstackx Collaborator Loading…
5
[Klaud Cold] Add H200 Qwen3.8-27B-FP8 vLLM TP1 native MTP 1k1k / 新增 H200 Qwen3.8-27B-FP8 vLLM TP1 原生 MTP 1k1k 配方 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3293 opened Sep 19, 2026 by functionstackx Collaborator Loading…
3
[Klaud Cold] Add H100 Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k / 新增 H100 Qwen3.8-27B-FP8 vLLM MTP 1k1k + 8k1k 配方 full-sweep-fail-fast qwen3.8-27b Qwen3.8-27B (bf16) vLLM TP1 DSpark 1k1k recipes and their eager variants
#3292 opened Sep 19, 2026 by functionstackx Collaborator Loading…
3
[AgentX GB200] enable Kimi-K3 power on the remaining recipes and re-sweep the full curve / 为剩余配方启用实测功耗并重跑完整曲线 ci-patchwork-waived priority Preempt other runs on this sweep's runners; restore them at the end (org members only) skip_queue
#3287 opened Sep 19, 2026 by edwingao28 Collaborator • Draft
3 of 5 tasks
1
ProTip! Exclude everything labeled bug with -label:bug.