Add TF weights

#3
by joaogante HF staff - opened

Model converted by the transformers' pt_to_tf CLI.

All converted model outputs and hidden layers were validated against its Pytorch counterpart. Maximum crossload output difference=1.335e-05; Maximum converted output difference=1.335e-05.

cc @patrickvonplaten [HF maintainer(s) for this repo]

(I see there is another TF weights PR, although there is a small mismatch on the weights. The difference could be explained by the position ids, which don't need to be loaded.)

Ok to merge for me - thanks @joaogante

joaogante changed pull request status to merged

I'll close my PR then.

Sign up or log in to comment