GitHub · Where software is built

Labels Milestones New issue

[Bug] Added some tokens to the tokenizer and added modules_to_save = ["embed_tokens", "lm_head"], MERGE MODEL PERFORMANCE IS BAD.

#3444

· diazr04 opened

on Oct 13, 2025

OOM-ing on Nvidia Jetson Orin Nano

#3443

· rlleshi opened

on Oct 13, 2025

[Bug] 微调qwe3-4B模型，内存占用将近60G，训练集只有7M，请教一下为什么？

#3442

· WZNoone opened

on Oct 13, 2025

[Bug] type object 'CompiledKernel' has no attribute 'launch_enter_hook'

#3441

· ATRI-Star opened

on Oct 13, 2025

[Bug] Getting Qwen3ForCausalLM.forward() got multiple values for argument 'input_ids' with GRPO

feature request

#3439

· adi-kmt opened

on Oct 11, 2025

CUDA out of memory Error

#3435

· NiklasWillecke opened

on Oct 10, 2025

[Bug] FastModel Doesn't Support MultiGPU Finetuning.

#3434

· Vinayyyy7 opened

on Oct 10, 2025

[Bug] AssertionError: No inf checks were recorded for this optimizer when finetune_language_layers=False

#3433

· humeedat opened

on Oct 10, 2025

trainer.train() stuck in pytorch inductor compilation after 100-724 steps

#3432

· julianghadially opened

on Oct 10, 2025

[Urgent Bug] Unsloth: No config file found

#3431

· silentgameshub opened

on Oct 10, 2025

[Feature] Need docker image for aarch64 platform

feature request

#3429

· xuancong84 opened

on Oct 10, 2025

[Bug] Adapter merging and saving are corrupted for Qwen3

#3428

· DmitryDiTy opened

on Oct 9, 2025