Adding _set_gradient_checkpointing for compatibility

#22
by vriveras - opened

Adding _set_gradient_checkpointing for compatibility when finetuning the model.

vriveras changed pull request title from Update modeling_mixformer_sequential.py to Adding _set_gradient_checkpointing for compatibility

Hi,

I find that the updated code for gradient checkpointing does not work in my case, i.e., the memory usage is not reduced.

I wonder whether this code is tested in practice.

Best regards,
Ziniu

Please fix would be very nice

require for fix @gugarosa

Microsoft org

Could you please re-update your file with the latest commit? As soon as the merge conflict is solved, I will merge this PR.

I have rebased the PR.

@gugarosa will you be able to merge this?

gugarosa changed pull request status to merged

Sign up or log in to comment