Skip to content

Fix masked parameters training in 1-bit Adam - #1278

Merged
conglongli merged 1 commit into
deepspeedai:masterfrom
DT6A:patch-1
Aug 5, 2021
Merged

Fix masked parameters training in 1-bit Adam#1278
conglongli merged 1 commit into
deepspeedai:masterfrom
DT6A:patch-1

Conversation

@DT6A

@DT6A DT6A commented Aug 5, 2021

Copy link
Copy Markdown
Contributor

Make add operation inplace. Without it momentum decays to zero and training has no effect on corresponding parameters

Make add operation inplace. Without it momentum decays to zero and training has no effect on corresponding parameters
@conglongli
conglongli merged commit 00320a9 into deepspeedai:master Aug 5, 2021
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