Finetuning

[asunaperisi]

It is great job.Thank you to share. I just want to ask how can I fine-tune your model with my own dataset? I saw you already added vicap.

I am using this code for fine-tuning and same vocab file that you provide.Naturally some of the tokens of my dataset is not included at vocab file. First I thought I can simply add the tokens to the vocab file. But there is a parameter named vocab_size. I updated this parameter properly. When I try to use your per-trained model, I am getting size mismatch error. Is there any way for fine-tuning without retrain the entire model.
Thank you.