search

ai-safety-foundation / sparse_autoencoder

Sparse Autoencoder for Mechanistic Interpretability
https://ai-safety-foundation.github.io/sparse_autoencoder/
MIT License
171 stars 39 forks source link

`poetry run python sparse_autoencoder/training_runs/gpt2.py` tries to write to directories outside the git repository #204

Closed JasonGross closed 5 months ago

JasonGross commented 6 months ago 
Traceback (most recent call last):
  File "/home/jason/sparse_autoencoder/sparse_autoencoder/train/sweep.py", line 255, in train
    run_training_pipeline(
  File "/home/jason/sparse_autoencoder/sparse_autoencoder/train/sweep.py", line 202, in run_training_pipeline
    checkpoint_path.mkdir(exist_ok=True)
  File "/opt/conda/envs/py3.11/lib/python3.11/pathlib.py", line 1116, in mkdir
    os.mkdir(self, mode)
PermissionError: [Errno 13] Permission denied: '../../.checkpoints'