Fine-tuning using pretrained model

[SebastianLarssonDTU]

I am trying to train the reader using one of the pretrained models, but the "optimizer.pth.tar" does not exist in the folder (in this case nq_reader_base). Is it possible to use one of the pretrained models for fine-tuning?

And is the eval_dataset strictly necessary for training the reader?

Thanks in advance,
Sebastian

[wentinghome]

Hi, I have the same question