In this image I don’t know the meaning of this total_flos. Also as a additional question, what is the difference between train/loss and train/train_loss?
loss vs. train_loss is at the bottom of this: [trainer] 'train_loss' different from 'loss'
flos is here (kinda): "total_flos" showing much bigger number than expected · Issue #15006 · huggingface/transformers · GitHub
1 Like