Search results

[Bug] Chat Template Not Applied Correctly to Recent Models

Status: Open.

#551 In sgl-project/SpecForge;

· Dogacel opened on Apr 30, 2026

High results fluctuation in repeating experiments with same setup

Status: Open.

#550 In sgl-project/SpecForge;

· Rachelcoll opened on Apr 28, 2026

Question about train dflash for a large target model

Status: Open.

#546 In sgl-project/SpecForge;

· dongyibo opened on Apr 27, 2026

[Feature] Replace --is-reasoning-model with unified --reasoning flag in regenerate_train_data.py

Status: Open.

#544 In sgl-project/SpecForge;

· msober opened on Apr 24, 2026

[Feature] Support for PEAGLE

Status: Open.

#541 In sgl-project/SpecForge;

· bhaktatejas922 opened on Apr 23, 2026

[Bug] There is a gap between the acceptance rates of training and inference

Status: Open.

#533 In sgl-project/SpecForge;

· xinanjiao opened on Apr 14, 2026

[Feature] Need offline training support for Dflash

Status: Open.

#521 In sgl-project/SpecForge;

· UltramanKuz opened on Apr 1, 2026

[Bug] qwen3.5-35b-a3b dflash

Status: Open.

#520 In sgl-project/SpecForge;

· Jim2016713 opened on Mar 31, 2026

[Bug] Dflash

Status: Open.

#501 In sgl-project/SpecForge;

· Jim2016713 opened on Mar 16, 2026

[Bug] with how mask_token_id is selected in the dflash script; it does not use the value configured in draft_config.dflash_config .

Status: Open.

#500 In sgl-project/SpecForge;

· dfiyoyo opened on Mar 13, 2026

[Feature] Plots using benchmarking result files

Status: Open.

#499 In sgl-project/SpecForge;

· MLKoz2 opened on Mar 12, 2026

[Feature] will support qwen3.5-mtp training and qwen3.5-eagle3 training?

Status: Open.

#497 In sgl-project/SpecForge;

· aidenchn opened on Mar 9, 2026