Repository navigation

All pull requests

Qualify ATOM cache reuse against profiling requests / 核对 ATOM 缓存复用的 profiling 请求范围
#4405chunfangamd opened Oct 22, 2026Collaborator
6
[AgentX] Update Muse Spark Open GB300 SGLang image and retune concurrency points / [AgentX] 更新 Muse Spark Open GB300 SGLang 镜像并调整并发点
#4404shyeh25 opened Oct 22, 2026Collaborator4 of 11 tasks
2
[Klaud Cold] Update MiniMax-M3 FP8 MI300X vLLM AgentX image to v0.32.0 / [Klaud Cold] 将 MiniMax-M3 FP8 MI300X vLLM AgentX 镜像更新至 v0.32.0
#4402Klaud-Cold opened Oct 22, 2026Collaborator
7
[NVIDIA][AgentX] GB300 DeepSeek-V4-Pro FP4 Dynamo+SGLang: PP8-prefill disagg and TP4/TP8 agg DSpark points / [NVIDIA][AgentX] GB300 DeepSeek-V4-Pro FP4 Dynamo+SGLang：PP8 prefill 分离式与 TP4/TP8 聚合 DSpark 配置点
#4401nvpohanh opened Oct 22, 2026Collaborator
6
[AgentX] Update Muse Spark Open B200 SGLang image and add c64 / [AgentX] 更新 Muse Spark Open B200 SGLang 镜像并新增 c64
#4400shyeh25 opened Oct 22, 2026Collaborator4 of 11 tasks
2
docs: clarify GB300 AgentX result provenance / 澄清 GB300 AgentX 结果来源与可比性
#4399Juntian777 opened Oct 22, 2026Collaborator
1
[NVIDIA][AgentX] GB200 GLM-5.2 NVFP4 Dynamo+SGLang: PP8-prefill disagg and TP4/TP8 agg points / GB200 GLM-5.2 NVFP4 Dynamo+SGLang：PP8 prefill 分离式与 TP4/TP8 聚合配置点
#4398nvpohanh opened Oct 22, 2026Collaborator
3
Fix ATOMesh KV-cache counters / 修复 ATOMesh KV-cache 计数
#4397chunfangamd opened Oct 22, 2026Collaborator
5
[Klaud Cold] Update Qwen3.5 FP8 B200 SGLang AgentX MTP image to v0.5.22-cu130 / [Klaud Cold] 将 Qwen3.5 FP8 B200 SGLang AgentX MTP 镜像更新至 v0.5.22-cu130
#4396Klaud-Cold opened Oct 22, 2026Collaborator
9
[Klaud Cold] Update MiniMax-M3 FP8 MI325X vLLM AgentX image to v0.32.0 / [Klaud Cold] 将 MiniMax-M3 FP8 MI325X vLLM AgentX 镜像更新至 v0.32.0
#4395Klaud-Cold opened Oct 22, 2026Collaborator
8
[Klaud Cold] Update GLM-5.2 FP4 B200 SGLang AgentX MTP image to v0.5.22-cu130 / [Klaud Cold] 将 GLM-5.2 FP4 B200 SGLang AgentX MTP 镜像更新至 v0.5.22-cu130
#4394Klaud-Cold opened Oct 22, 2026Collaborator
9
[AMD][MI355X] Kimi-K3 vLLM AgentX nightly df6603f0 with latency_specialized decode at c1/c2 / [AMD][MI355X] Kimi-K3 vLLM AgentX 更新镜像至 nightly df6603f0，c1/c2 启用 latency_specialized decode
#4393seungrokj opened Oct 22, 2026Collaborator
6
Refresh Kimi-K3 PD on ROCm nightly / 更新 Kimi-K3 PD nightly 配置
#4391YukioZzz opened Oct 22, 2026Collaborator7 of 11 tasks
5
[Klaud Cold] Update DeepSeek-V4.1-Flash FP4 H100 vLLM AgentX image to nightly-8e2beaa3 / [Klaud Cold] 将 DeepSeek-V4.1-Flash FP4 H100 vLLM AgentX 镜像更新至 nightly-8e2beaa3
#4388Klaud-Cold opened Oct 22, 2026Collaborator
7
[NVIDIA][AgentX] MiniMax-M3 B200 vLLM digest pin, TP4 DRAM-offload tuning and capped-active c10 entries / [NVIDIA][AgentX] MiniMax-M3 B200 vLLM 按摘要固定镜像、调优 TP4 DRAM 卸载并新增限制活跃序列的 c10 配置
#4382hshrivastava-droid opened Oct 21, 2026Collaborator6 of 11 tasks
8
[NVIDIA][AgentX] Kimi-K3 B300 FP4 vLLM: DSpark 7/5/3 at conc 1-40 / [NVIDIA][AgentX] Kimi-K3 B300 FP4 vLLM：并发 1–40 使用 DSpark 7/5/3
#4381hshrivastava-droid opened Oct 21, 2026Collaborator6 of 11 tasks
5
[AMD][Qwen3.5] [Perf] Recover MI355X FP8 AgentX concurrency 80 with HiCache write_through on sglang-rocm v0.5.22-20261020
#4374yichiche opened Oct 21, 2026Collaborator
3
Refine Kimi-K3 GB300 AgentX points / 优化 Kimi-K3 GB300 AgentX 点位
#4373wzhao18 opened Oct 21, 2026Collaborator
5
CollectiveX: pin nccl-ep to nccl-extensions 0.2.0.dev20261013 / CollectiveX：将 nccl-ep 固定到 nccl-extensions 0.2.0.dev20261013
#4372Oseltamivir opened Oct 21, 2026Collaborator
CollectiveX: run flashinfer-ep on FlashInfer 0.7.1 / CollectiveX：让 flashinfer-ep 使用 FlashInfer 0.7.1
#4371Oseltamivir opened Oct 21, 2026Collaborator
3
[AgentX] Enable unified memory for B200 Muse Spark / [AgentX] 为 B200 Muse Spark 启用 unified memory
#4366Fridge003 opened Oct 21, 2026Collaborator6 of 11 tasks
2
MiniMax-M3 NVFP4 B300 fp4 vLLM AgentX: tuned config-only curve on nightly-ac68c308 / MiniMax-M3 NVFP4 B300 fp4 vLLM AgentX：基于 nightly-ac68c308 的纯配置调优曲线
#4357hshrivastava-droid opened Oct 20, 2026Collaborator5 of 10 tasks
1
perf(qwen3.5): move GB300 TRT AgentX to TRT-LLM 1.3.0a0.dev202610200641 / perf(qwen3.5): GB300 TRT AgentX 切换到 TRT-LLM 1.3.0a0.dev202610200641
#4356richardhuo-nv opened Oct 20, 2026Collaborator
14
[AgentX] Add GB300 DeepSeek-V4.1-Flash disagg 10P1D c8400 point
#4352cquil11 opened Oct 20, 2026Collaborator
3
Add B300 Dynamo+SGLang AgentX configs for DeepSeek V4 with Mooncake EFA KV transfer / 添加使用 Mooncake EFA KV 传输的 DeepSeek V4 B300 Dynamo+SGLang AgentX 配置
#4346nvpohanh opened Oct 20, 2026Collaborator7 of 11 tasks
18

Pull requests