GitHub · Where software is built

Labels Milestones New issue

[Bug] Can't finetune LiquidAI LFM2.5-VL-1.6B with vision

#3938

· urbanspr1nter opened

on Jan 26, 2026

[Bug] embeddinggemma saving issue

feature request

#3932

· aheluc opened

on Jan 25, 2026

AttributeError: 'GRPOConfig' object has no attribute 'unsloth_grpo_mini_batch'

#3931

· left0ver opened

on Jan 25, 2026

[Feature] GDPO: New GRPO modification for multi-reward RL

feature request

#3930

· kabachuha opened

on Jan 25, 2026

[Bug] Cannot save model with gpt-oss when target_modules include "embed_tokens", "lm_head" (e.g., for CPT)

#3926

· davedgd opened

on Jan 23, 2026

[Bug] model.language_model.layers.25.mlp.up_proj.SCB'} do not match! Qwen 3 vl 4b

#3925

· xellDart opened

on Jan 22, 2026

[Bug] RuntimeError: Triton Error [CUDA]: an illegal memory access was encountered with PRO RTX6000

#3921

· camposs1979 opened

on Jan 22, 2026

Isolate GGUF exports into a dedicated <code>_gguf</code> directory

#3919

· RektPunk opened

on Jan 22, 2026

[Bug] Multi GPU (DDP) is not working in any setting. Balanced only on Magistral?

#3915

· saurabhbikram opened

on Jan 21, 2026

Qwen3_VL_(8B)-Vision-GRPO.ipynb failed when use multiple GPUs

#3914

· dajuanmaoa opened

on Jan 21, 2026

[Bug] RuntimeError: Engine core initialization failed. See root cause above. Failed core proc(s): {}

#3911

· saadraqib opened

on Jan 20, 2026

Support for GRPO multi-GPU training with Qwen2.5?

#3901

· ChengHu98 opened

on Jan 17, 2026