There is a way to understand all the parameters, loss functions and stuff, used during the training of the model?
I’m doing Fine-Tuning of the ViT on a dataset called “Stanford dog Dataset” for fine-grained purpose and I’m trying now to use it in Keras (I’m so sorry
) because I need the file in “.h5” format.
With HF fine-tuning I obtain like 99.6% of accuracy on TR and 99.1% in Val.
When I try the training with the other library, the accuracy is 82% Val.
Can someone help with that?