Fine tuning a new TF model
Hi Yih-Dar,
I am trying to fine-tune a new TF model on top of your published model. The colab I have experimented is https://colab.research.google.com/github/patrickvonplaten/notebooks/blob/master/BERT2BERT_for_CNN_Dailymail.ipynb It is PyTorch. There is a gap between this and TensorFlow, particularly Seq2SeqTrainer() is not available in TF. I wonder if you can share any suggestions or code pointers since you have worked on a TensorFlow model. Thanks a lot!
Hi @tiktok-legacy Probably you no longer need the information anymore (sorry missed this thread).
Quoted from the TF person in our team:
TFTrainer is very old and completely deprecated now, and we don't support it anymore. We generally advise people to just use the Keras API for TF.
More information: https://github.com/huggingface/transformers/issues/26632#issuecomment-1750960315
If you need more information about the TF/Keras documentation, let me know.