Skip to content

Add files via upload#4

Open
frankliuyc wants to merge 2 commits into
mainfrom
frankliuyc-patch-2
Open

Add files via upload#4
frankliuyc wants to merge 2 commits into
mainfrom
frankliuyc-patch-2

Conversation

@frankliuyc
Copy link
Copy Markdown
Collaborator

#1
filter_state_dict(model, stashed_dict, fill_tensor=False) will now reshape branched layer weight/bias to accommodate additional branches when fill_tensor=True

@sjgosai
Copy link
Copy Markdown
Owner

sjgosai commented Jul 21, 2022

@frankliuyc One suggestion. You've coded in an implicit assumption that tensors that aren't the right shape should be modified in the 0th dimension. I would add in some kind of check that every other dim has the same size (except dim=0) before filling. Otherwise, looks great! Almost ready to pull.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants