-
Notifications
You must be signed in to change notification settings - Fork 24.2k
Bf16 fused adam(W) #147653
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.
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
base: gh/janeyx99/225/base
Are you sure you want to change the base?
Bf16 fused adam(W) #147653
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/147653
Note: Links to docs will display an error until the docs builds have been completed. ❌ 12 New Failures, 1 Cancelled JobAs of commit 6a74269 with merge base 83bb921 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
Looks like this PR hasn't been updated in a while so we're going to go ahead and mark this as |
Many things do work!
Some things do not:
amsgrad does not work
some checks are removed (not critical imo, but def less safe) to let multidevice work
adam + adamw should work
tensor lr would work if it's on cpu
the float, float, bf16, bf16 pattern is specialized, other mixed precision will not work
Stack from ghstack (oldest at bottom):