GitHub · Where software is built

Labels Milestones New issue

[Bug] SFTTrainer isssues

#3657

· Hert4 opened

on Nov 30, 2025

[Bug] Gemma 3n - maximum recursion depth exceeded

#3650

· Tacx79 opened

on Nov 28, 2025

[Bug] Can't load tokenizer through FastModel.from_pretrained while loading local cached model

#3647

· YunkaiXiao opened

on Nov 26, 2025

[Bug] Method _from_pretrained in Class PreTrainedTokenizerBase of tokenization_utils_base.py is not robust

#3646

· chyupen opened

on Nov 26, 2025

[Feature] Support for Search-r1

feature request

#3642

· dipta007 opened

on Nov 25, 2025

[Feature] More granular quantization options for VL models when using FastVisionModel

feature request

#3639

· leecming82 opened

on Nov 25, 2025

[Feature] Qwen3-omni TTS Voice Cloning Support

feature request

#3636

· abrar360 opened

on Nov 25, 2025

[Bug] save_pretrained_merged("merged_16bit") re-downloads FP16 weights into output .cache instead of using HF cache

#3633

· shimmyshimmer opened

on Nov 24, 2025

UnslothVisionDataCollator() with streaming datasets.

#3630

· poilly54 opened

on Nov 24, 2025

[Bug] Models already trained - getting stuck at training run

#3627

· Maxtheconquerer opened

on Nov 21, 2025

[Bug] num_generations seems to be ignored in GRPO rollout

#3626

· yuxizhe opened

on Nov 21, 2025

[Bug] Some tokenizers broken when using <code>FastLanguageModel</code>

fixed - pending confirmation

#3624

· BramVanroy opened

on Nov 20, 2025