Closed MayDomine closed 5 months ago
Scale formula changes from loss = loss / dp_size to loss / (dp_size * tp_size).
loss = loss / dp_size
loss / (dp_size * tp_size)
Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce.
Fix scale loss logic
Description
Scale formula changes from
loss = loss / dp_size
toloss / (dp_size * tp_size)
.Type of Change
How Has This Been Tested?
Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce.
Checklist