[Bug] Private Dataset Metadata Request Does not pass the hf token

[HadiSDev]

Note: Please do not remove the questions. Answer beside them.

1. Did you update? pip install --upgrade unsloth unsloth_zoo
   Yes
2. Colab or Kaggle or local / cloud
   RunPod + Docker image latest tag
3. Number GPUs used, use nvidia-smi
   1
4. Which notebook? Please link!
   No Notebook
5. Which Unsloth version, TRL version, transformers version, PyTorch version?
   unsloth_zoo-2026.4.6 | I just use the unsloth:latest tag
6. Which trainer? SFTTrainer, GRPOTrainer etc
   Not relevant as it is before training

If I use private datasets where I need to pass a hf_token the metadata does not load (what splits exist or length). However, the data itself loads fine and I can train as well. If I switch it to public I no longer have the issue and I can see all my splits. Seems like the hf token was only fetched for dataset preview but not for the metadata.

[Imagineer99]

Seems plausible. Preview/training and subset metadata currently go through different paths, so it could be possible for private dataset loading to work while metadata fails.