Morning

  • Asked Peter to let me join the AI training team.
  • I transferred from using to actually building an AI model.
  • After following the instruction, I encountered “AssertionError” in Python, and stucked.

    assert vocab_file is not None AssertionError

Afternoon

  • Found someone who encountered exactly the same problem on Google.
  • The problem was solved by someone.
  • He says that I was lack of a file “ice_text.model”.
  • Consequently, I went to the remote repository’s website to checkout.
  • I found that all the LFS-managed big files were not cloned.
  • In other words, the clone was not complete.
  • Solved by initializing LFS of git:

    git lfs install

  • Downloading large files always need patience…
  • Encountered another error:

    AssertionError: Torch not compiled with CUDA enabled

  • Workaround: if I run the model with CPU, it goes fine, but it’s too slow.
  • Real Solution: Reinstalling PyTorch with CUDA Support
  • Resolved by going to the PyTorch official website, pytorch.org, and ran the following command;

    pip3 install torch torchvision torchaudio –index-url https://download.pytorch.org/whl/cu117

  • Uploaded all the previous working including Python and JS to the DaoCloud gitlab repository.
  • Lesson learned: pull from the origin before working on a github project.