Skip to content

Support the maximize parameter for adam when dtype is torch.half #35

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 4 commits into from
Jul 8, 2022
Merged

Support the maximize parameter for adam when dtype is torch.half #35

merged 4 commits into from
Jul 8, 2022

Conversation

alphaGem
Copy link

@alphaGem alphaGem commented Jul 3, 2022

Related to issue34

a710128
a710128 requested changes Jul 4, 2022
View reviewed changes
@a710128
Copy link
Contributor

a710128 commented Jul 4, 2022

  1. AdamOffload needs to be updated.
    https://github.com/OpenBMB/BMTrain/blob/0.1.7.post1/bmtrain/optim/adam_offload.py#L131-L172
  2. The maximize parameter of torch.float also needs to be supported.
  3. A bug: grad was not passed to the adam function.

@a710128 a710128 linked an issue Jul 4, 2022 that may be closed by this pull request
[Feature] Supporting the maximize parameter of Adam optimizer when dtype is torch.half #34
Closed
@alphaGem
Copy link
Author

alphaGem commented Jul 5, 2022

Update:

  1. Adam Offload now updated.
  2. Maximize parameter of torch.float now supported.
  3. grad now passed to the adam function.

@a710128 a710128 merged commit e6ba031 into OpenBMB:main Jul 8, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@a710128 a710128 a710128 approved these changes

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.

[Feature] Supporting the maximize parameter of Adam optimizer when dtype is torch.half
2 participants
@alphaGem @a710128