GitHub · Where software is built

Labels Milestones New issue

[Feature] Create docs for exporting to <code>ONNX</code> format for webgpu inference support

feature request

#3808

· homanp opened

on Dec 31, 2025

Fail to finetune Qwen3-VL-30B-A3B-Instruct

#3807

· justStarG opened

on Dec 30, 2025

Issues fine-tuning VibeVoice with Unsloth (Transformers pinned, errors & GPU requirements)

#3805

· Diksha06122 opened

on Dec 30, 2025

[Bug] Unsloth + vLLM fast_generate() ignores LoRA on Gemma 3 270M

#3802

· Zajaczkowskim opened

on Dec 29, 2025

[Bug] NameError: name 'psutil' is not defined

#3800

· Romiroz opened

on Dec 29, 2025

fix: inputs_embeds ignored when input_ids is not None in _fast_prepare_inputs_for_generation

#3798

· majiayu000 opened

on Dec 29, 2025

[Feature Request]: Support for T5/ByT5 (Encoder-Decoder) Architecture

feature request

#3797

· shelterwff-byte opened

on Dec 29, 2025

[Bug] Cannot Load Qwen3Guard

#3791

· Calandracas606 opened

on Dec 29, 2025

Support for TRL GRPOTrainer with tools= argument (requires transformers 5.0)

#3786

· austinmw opened

on Dec 26, 2025

[Bug] llava1.5-7b-hf ValueError: Image features and image tokens do not match: tokens: 575, features 2359296

#3783

· Zuozhuo opened

on Dec 26, 2025

[Bug] fast rope backpropagation broken for zero strided tensors

#3781

· f14-bertolotti opened

on Dec 26, 2025

[Bug] ValueError: passed inputs_embeds to .generate() when using LlamaForCausalLM with Unsloth

#3779

· Pioneer-Weirdo opened

on Dec 25, 2025