79347858

Date: 2025-01-11 10:06:05
Score: 2.5
Natty:
Report link

Thanks a lot for sharing this knowledge, not just this. Trainer class of hugging face has a lot of issues. Their evaluation loop involves first storing all the predictions and then computing the metrics. For NLP, this can take up a lot of memory and one may even face out of memory errors. It is best to write your own pytorch training loop.

Reasons:
  • Blacklisted phrase (0.5): Thanks
  • No code block (0.5):
  • Single line (0.5):
  • Low reputation (1):
Posted by: Radhika Desai