Search results

[Bug] Severe Decoding Throughput Regression with FP4 KV Cache (~90 tokens/s vs ~5000 tokens/s)

Status: Open.

#17365 In sgl-project/sglang;

· jindajia opened on Jan 19, 2026

[Bug] MoE Models Fail on Ascend 910B During CUDA Graph Capture

Status: Open.

#17362 In sgl-project/sglang;

· darthvader58 opened on Jan 19, 2026

[Bug] Quen-3-VL-235B won't start with pp-size 2 (KeyError: 'model.layers.15.mlp.experts.w2_weight')

Status: Open.

#17359 In sgl-project/sglang;

· Swipe4057 opened on Jan 19, 2026

[Bug] Generating queries without SO stops prematurely

Status: Open.

#17357 In sgl-project/sglang;

· Swipe4057 opened on Jan 19, 2026

[Bug] spec v2 eagle3 logprobs error

speculative-decoding

Status: Open.

#17349 In sgl-project/sglang;

· ispobock opened on Jan 19, 2026

[Bug] DeepSeek-V3.2 SWE-bench Verified score significantly lower than official results

Status: Open.

#17348 In sgl-project/sglang;

· ovowei opened on Jan 19, 2026

increase in GPU memory

Status: Open.

#17341 In sgl-project/sglang;

· ZhanguoChang opened on Jan 19, 2026

How to build sgl-kernel from source?

Status: Open.

#17333 In sgl-project/sglang;

· LifuWang-66 opened on Jan 19, 2026

[Bug] Output token count abnormally increases with Qwen3-Next when --enable-piecewise-cuda-graph is set

Status: Open.

#17330 In sgl-project/sglang;

· lw9527 opened on Jan 19, 2026

[Feature] [Tracking] Enable RDMA weight transfer for RL use cases

Status: Open.

#17311 In sgl-project/sglang;

· JD-ETH opened on Jan 18, 2026

[Bug][Diffusion][Wan] STA (and other sparse attention) is applied to cross‑attention in Wan; should be self‑attn‑only?

Status: Open.

#17303 In sgl-project/sglang;

· tie-pilot-qxw opened on Jan 18, 2026

[Bug] AttributeError: type object 'PassConfigKey' has no attribute 'TL_DISABLE_FAST_MATH'. Did you mean: 'TL_ENABLE_FAST_MATH'?

Status: Open.

#17298 In sgl-project/sglang;

· magicdy opened on Jan 18, 2026