Ungraded Lab: Question Answering with HuggingFace 2

Course Q&A Natural Language Processing NLP with Attention Models
1

Based on the discussion of the lab here:

For this lab, you will use the same model from the question-answering pipeline that you loaded before. But, if you want, you can uncomment the cell below to load the pre-trained model without accessing the pipeline parameters.

You will set the base model layers fixed so the training process doesn’t take too long. To do so, you have to set the requires_grad parameter for the base_model layers to False .

However, I can’t see the relevant parameter settings in the code.

Can you explain a bit?

2

@tyyim Even after a month no one explained it. I tried my level best but in vain. Could you figure it out? Let me know you did.

3

It appears there is no cell to uncomment in the notebook, so the statement about requires_grad has no relevance to the code that is in fact provided. I have made a note to people at the backend to remove the relevant passages.