Search results

[Bug] DFlash block mask currently relies on Flex Attention only — request for SDPA / generic attention backend support

Status: Open.

#474 In sgl-project/SpecForge;

· baifanxxx opened on Feb 26, 2026

[Bug] dflash cross block data leak

Status: Open.

#470 In sgl-project/SpecForge;

· TechxGenus opened on Feb 13, 2026

[Feature] What is the upper limit to the performance improvement of Dflash training as the step increases?

Status: Open.

#469 In sgl-project/SpecForge;

· 5SSjw opened on Feb 13, 2026

[Bug] Abnormal memory usage and Out-of-Memory in eagle3 training.

Status: Open.

#466 In sgl-project/SpecForge;

· coco-alen opened on Feb 12, 2026

[Feature] [RFC] DFlash Training Adaptation for Qwen VL Models

Status: Open.

#461 In sgl-project/SpecForge;

· ggg-s opened on Feb 5, 2026

EAGLE for GPT-OSS-20B

Status: Open.

#459 In sgl-project/SpecForge;

· yanzi-elastix opened on Feb 3, 2026

[Bug] flex_attention BlockMask incorrectly created with H=num_heads

Status: Open.

#452 In sgl-project/SpecForge;

· Ximingwang-09 opened on Jan 27, 2026

[Feature] [RFC] Add a step-disaggregated draft model for speculative decoding, PRISM

Status: Open.

#444 In sgl-project/SpecForge;

· Akemiiii opened on Jan 21, 2026

[Regression] requirements.txt is empty

Status: Open.

#432 In sgl-project/SpecForge;

· Dogacel opened on Jan 18, 2026

[Feature] Optimization of Storage Efficiency for Hidden States in Large-Scale Models

Status: Open.

#426 In sgl-project/SpecForge;

· shelterwff-byte opened on Jan 13, 2026

[Bug] BUILD_DATASET_NUM_PROC>1 and the data length is greater than 32768 will block mapping.

Status: Open.

#422 In sgl-project/SpecForge;

· jiapingW opened on Jan 12, 2026

[Feature] Improve multi-layer decoder of custom backends

Status: Open.

#418 In sgl-project/SpecForge;

· xiaoxi-s opened on Jan 11, 2026