08 > NLP > NLP: Attaching a Head - output.shape or output.logits.shape?

p.hanel

See the below bolded text. It was otherwise difficult to see.

This is the course:

I did try running but got:

I think the fix is adding '.logits':

Please let me know if it was just the case of an update changing the parameter or if I am displaying something else not intended. Thank you.

dvgodoy

Hi @p.hanel ,

You're right, we're supposed to read the logits. Thank you for pointing it out.

It should read:

model_with_head.eval()
output = model_with_head(model_input)
output, output.logits.shape

Best,
Daniel

dharmon

Hi @p.hanel ,

Thank you again for your feedback. The course has been updated with the changes.

Kind Regards,

David
Linux Foundation Education Team

p.hanel

Thank you for the quick responses @dharmon and @dvgodoy 😊.