modernBERT training learning rate=0 and validation_loss=nan
#66
by
devals9
- opened
I have latest version of transformers, torch installed. I train the model using AutoModelForSequenceClassification but I keep getting learning rate as 0 and validation loss as nan. I have not installed flash-attn
What are the correct version of packages that I should install ?
Any help is appreciated.
This comment has been hidden