Search results

[Bug] silu_and_mul and other activation kernels crash with cudaErrorInvalidConfiguration when num_tokens=0

Status: Open.

#23609 In sgl-project/sglang;

· imReese opened on Apr 24, 2026

[Bug] Multi-node deployment crashes with collective sequence number mismatch in scheduler initialization (Gemma4)

Status: Open.

#23603 In sgl-project/sglang;

· solitude-alive opened on Apr 24, 2026

DeepSeek V4 Roadmap

deepseek

high priority

roadmap

Status: Open.

#23602 In sgl-project/sglang;

· fzyzcjy opened on Apr 24, 2026

WIP

Status: Open.

#23598 In sgl-project/sglang;

· iforgetmyname opened on Apr 24, 2026

[Bug] Hy3-preview cuda-graph crash on AMD MI300X/MI355X due to AITER custom all-reduce stream invalidation

Status: Open.

#23580 In sgl-project/sglang;

· andyluo7 opened on Apr 23, 2026

Session Reaper Mid-Decode Race Condition & /v1/completions Parameter Drop

Status: Open.

#23579 In sgl-project/sglang;

· glaziermag opened on Apr 23, 2026

[diffusion] model: support ar + diffusion hybrid model

Status: Open.

#23578 In sgl-project/sglang;

· mickqian opened on Apr 23, 2026

[Bug] PD disaggregation decode worker crashes (CUDA IndexKernel “index out of bounds”) with Qwen/Qwen3.6-27B using NIXL backend (H100, sglang 0.5.10.post1)

Status: Open.

#23574 In sgl-project/sglang;

· Bihan opened on Apr 23, 2026

[Feature] Refactor decode-side KV offload to reuse HostPoolGroup / HybridCacheController and enable NSA support

Status: Open.

#23571 In sgl-project/sglang;

· Sebastian-dong opened on Apr 23, 2026

[NPU]LLaDA2.1-flash is much slower on Ascend 910B than on H20 in SGLang

Status: Open.

#23565 In sgl-project/sglang;

· ApsarasX opened on Apr 23, 2026

[Bug] Kimi K2.6 DEP8 produces garbage output but DP8 works fine

Status: Open.

#23554 In sgl-project/sglang;

· mmangkad opened on Apr 23, 2026

[Bug] 0.5.10 PD disaggregated nixl-cu13 error：NIXL_ERR_NOT_FOUND

Status: Open.

#23551 In sgl-project/sglang;

· DingYinfan opened on Apr 23, 2026