Hi. The finetuning results has these output column headers.
step
elapsed_tokens
elapsed_examples
training_loss
training_sequence_accuracy
training_token_accuracy
validation_loss
validation_sequence_accuracy
validation_token_accuracy
I created graphs of step & training accuracy and Validation accuracy to ensure that my model is running correctly. What other criteria can help me understand if my model is work fine?
P.S. - I am looking for intrinsic evaluation rather than extrinsic evaluation.