Search results

[Docs] Update speculative decoding document

documentationImprovements or additions to documentationImprovements or additions to documentation

good second issueGood for second time and more contributorsGood for second time and more contributors

Status: Open.

#18268 In sgl-project/sglang;

· b8zhong opened on Feb 4, 2026

[Docs] Add doc for piecewise CUDA graph

documentationImprovements or additions to documentationImprovements or additions to documentation

good second issueGood for second time and more contributorsGood for second time and more contributors

Status: Open.

#18267 In sgl-project/sglang;

· b8zhong opened on Feb 4, 2026

[Feature] Add π0/π0-FAST VLA model inference support

Status: Open.

#18266 In sgl-project/sglang;

· yicwang opened on Feb 4, 2026

[Feature] Integrate FA4 SM90 Decode

good second issueGood for second time and more contributorsGood for second time and more contributors

hopperSM90SM90

Status: Open.

#18265 In sgl-project/sglang;

· b8zhong opened on Feb 4, 2026

[Docs] Verify FA4 in attention backend compatibility

documentationImprovements or additions to documentationImprovements or additions to documentation

good second issueGood for second time and more contributorsGood for second time and more contributors

Status: Open.

#18264 In sgl-project/sglang;

· b8zhong opened on Feb 4, 2026

[Bug] [AITER attention] OOM errors as resolved <code>max_total_num_tokens</code> does not take into account the memory requirements from attention backends

Status: Open.

#18262 In sgl-project/sglang;

· fxmarty-amd opened on Feb 4, 2026

[ModelOpt] [Good issue] [docs] Explain how to use nvidia modelopt checkpoints

good second issueGood for second time and more contributorsGood for second time and more contributors

Status: Open.

#18261 In sgl-project/sglang;

· vincentzed opened on Feb 4, 2026

[Feature] [ModelOpt] Loader doesn't support MXFP8

Status: Open.

#18258 In sgl-project/sglang;

· vincentzed opened on Feb 4, 2026

[Bug] API key in service logs

Status: Open.

#18256 In sgl-project/sglang;

· Swipe4057 opened on Feb 4, 2026

[Bug] GLM-4.7 EAGLE speculative decoding generates ~35% fewer tokens than max_tokens (draft tokens counted toward limit)

Status: Open.

#18251 In sgl-project/sglang;

· prathamk-tw opened on Feb 4, 2026

[Roadmap-HiCache]: Unified Storage Framework for SGLang Across Diverse Scenarios

hicacheHierarchical Caching for SGLangHierarchical Caching for SGLang

Status: Open.

#18239 In sgl-project/sglang;

· hzh0425 opened on Feb 4, 2026

Complexity in model layers and communication kernels

Status: Open.

#18237 In sgl-project/sglang;

· nvjullin opened on Feb 4, 2026