e-p-armstrong / e-p-armstrong/augmentoolkit
Enhancement — Adding GPU support to classifier training
Open
enhancement
good first issue
- Dominant language
- Python
- Stars
- 1.9k
- Forks
- 248
- PR merge metrics
- No merged PRs in 30d
Description
Problem description: I had to add a use_cpu=True flag to the TrainingArguments in the train_classifier function in steps.py, to get around an MPS issue that seemed to only happen in newer versions of transformers.
This will make classiifier training slower on decent machines.
Someone who knows or has time to look into transformers code — I'd welcome a PR that fixes this!
Contributor guide
No contributing guide indexed for this repository
Assessment
This issue has not been assessed yet.