Search results

Does the current Qwen3-VL (or Qwen3-VL-MoE) officially support TBO?

Status: Open.

#14903 In sgl-project/sglang;

· jerry-dream-fu opened on Dec 11, 2025

Default system prompt for DeepSeek-V3.2 Thinking can hinder the model performance

Status: Open.

#14902 In sgl-project/sglang;

· an1006634493 opened on Dec 11, 2025

[Bug] deepep not found on deepseek v3.2 using sgl 0.5.6 + EP8

Status: Open.

#14896 In sgl-project/sglang;

· xwpaul3 opened on Dec 11, 2025

When configuring PD Disaggregation, it gets stuck during memory allocation, especially when used in conjunction with Mooncake.

Status: Open.

#14885 In sgl-project/sglang;

· lululu-1997 opened on Dec 11, 2025

When configuring PD decomposition with 2p2d, the second prefiller fails to start properly.

Status: Open.

#14882 In sgl-project/sglang;

· lululu-1997 opened on Dec 11, 2025

[Bug] <code>AddReqResult.NO_TOKEN</code> sets <code>batch_is_full = True</code> too late to trigger preemption

Status: Open.

#14872 In sgl-project/sglang;

· Tuanshu opened on Dec 11, 2025

How to train vicuna EAGLE3 model?

Status: Open.

#14868 In sgl-project/sglang;

· Sylvan820 opened on Dec 11, 2025

Throughput degradation on Qwen3-30B-A3B with EAGLE3

Status: Open.

#14824 In sgl-project/sglang;

· Zzsf11 opened on Dec 10, 2025

[Bug] Grammar not working without reasoning (v0.5.6)

Status: Open.

#14821 In sgl-project/sglang;

· Emilmeister opened on Dec 10, 2025

[Bug] Performance regression with fa3 fp8 attention, worse than bf16 fa3.

Status: Open.

#14816 In sgl-project/sglang;

· jianfei-wangg opened on Dec 10, 2025

[Bug] RTX 5090, the attention_backend is automatically set to 'trtllm_mha', but a ValueError is raised during SM version detection.

Status: Open.

#14814 In sgl-project/sglang;

· sufeng-buaa opened on Dec 10, 2025

[Bug] Unstable dispatch time consumption of DeepSeek-V3.1 with PD disaggregation

Status: Open.

#14813 In sgl-project/sglang;

· Yi-sir opened on Dec 10, 2025