Search results

[Roadmap] Adaptive Speculative Decoding Roadmap

Status: Open.

#23705 In sgl-project/sglang;

· alphabetc1 opened on Apr 25, 2026

[Bug] [Hisparse] Hisparse CPU Memory Allocation Checker

Status: Open.

#23704 In sgl-project/sglang;

· xz-keg opened on Apr 25, 2026

[Bug] Qwen3.6-27B-FP8 (dense): FP8 weight_scale_inv silently dropped → garbage output (gate_gate_up_proj loop bug in qwen3_5.py)

Status: Open.

#23687 In sgl-project/sglang;

· gucasbrg opened on Apr 25, 2026

[sgl-kernel] per_token_group_quant_8bit_v2 masked path wastes ~90% of grid.y wave when per-expert token count is small

Status: Open.

#23683 In sgl-project/sglang;

· leonardHONG opened on Apr 25, 2026

[Bug] runner freezing for kimi-k2.6 model for some inputs

Status: Open.

#23676 In sgl-project/sglang;

· xhejtman opened on Apr 24, 2026

[Bug] PauseGeneration in PD-disaggregated mode: retract has no path back to prefill, and disagg event loops don't honor _engine_paused``

Status: Open.

#23668 In sgl-project/sglang;

· ByronHsu opened on Apr 24, 2026

[Tracker] DSV4 Performance Optimization Checklist

deepseek

nvidia

performance

Status: Open.

#23666 In sgl-project/sglang;

· nvpohanh opened on Apr 24, 2026

/v1/responses: previous_response_id injects system instructions instead of assistant output in conversation history

Status: Open.

#23662 In sgl-project/sglang;

· glaziermag opened on Apr 24, 2026

[Feature] HiCache Support for SWA (Gemma and Mistral)

Status: Open.

#23659 In sgl-project/sglang;

· baoskee opened on Apr 24, 2026

[Bug] DeepSeek-V4 compressed attention backend: no SM120 fallback for Lightning Indexer

Status: Open.

#23657 In sgl-project/sglang;

· rs-ipps opened on Apr 24, 2026

[Bug] [NPU] Qwen3.5-35B-A3B missing 1 required positional argument: 'interpolation'

Status: Open.

#23656 In sgl-project/sglang;

· OrangeRedeng opened on Apr 24, 2026

[Feature] [NPU] not supported prefix cache for Qwen3.5

Status: Open.

#23652 In sgl-project/sglang;

· LilyPankratova opened on Apr 24, 2026