Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.
And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.
Make pretrained models download path configurable
See original GitHub issueIs your feature request related to a problem? Please describe.
Directory where pretrained model is downloaded is hardcoded to
cache_dir = Path.joinpath(Path.home(), f'.cache/torch/NeMo/NeMo_{nemo.__version__}/{filename[:-5]}') in https://github.com/NVIDIA/NeMo/blob/main/nemo/core/classes/common.py#L652.
So model download fails in environments where $HOME is not writeable.
Describe the solution you’d like
Provide option to specify path to download pretrained model. May be via environment variable or a parameter?
Describe alternatives you’ve considered
To unblock myself, I have currently created a change in my fork to specify download path via environment variable. https://github.com/jarivk/NeMo/commit/f2e1c57e400b71ddcb0c76499c0cd7f29893ba9c
Additional context
None
Issue Analytics
- State:
- Created 2 years ago
- Comments:5
Top Results From Across the Web
Top Related Medium Post
Top Related StackOverflow Question
Troubleshoot Live Code
Top Related Reddit Thread
Top Related Hackernoon Post
Top Related Tweet
Top Related Dev.to Post
Top Related Hashnode Post
Hmm true it’s burried in documentation. You can use NEMO_CACHE_DIR=“SOME PATH” in your environment env (maybe bashrc or similar)
Hello! I understand that we can now change the location of default cache directory where the pre-trained models are downloaded to, is that correct? I can’t find how to do that in the documentation, could you help me with that?