1. Did you update? pip install --upgrade unsloth unsloth_zoo
   yes
2. Colab or Kaggle or local / cloud
   Colab
3. Number GPUs used, use nvidia-smi
   A100 80G on Colab
4. Which notebook? Please link!
   https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt_oss_(20B)_500K_Context_Fine_tuning.ipynb
5. Which Unsloth version, TRL version, transformers version, PyTorch version?
   official notebook
6. Which trainer? SFTTrainer, GRPOTrainer etc```python
   SFTTrainer

Put Minimal code to reproduce error here ###Remove Hugging Face token###

The official notebook contains a saved successful 500K run, but rerunning it now on Colab fails immediately at trainer.train() with CUDA out of memory. Tried to allocate 182.28 GiB on H100 and V100 80GB.

It's simply reproducible with official notebook..

Could you check it?

https://unsloth.ai/docs/blog/500k-context-length-fine-tuning

https://colab.research.google.com/github/unslothai/notebooks/blob/main/nb/gpt_oss_(20B)_500K_Context_Fine_tuning.ipynb

🦥 You can also ask via our Reddit page: https://reddit.com/r/unsloth/