DeepSpeed
set the default to use set_to_none for clearing gradients in BF16 optimizer.
#5434
Merged

set the default to use set_to_none for clearing gradients in BF16 optimizer. #5434

loadams merged 8 commits into deepspeedai:master from inkcherry:fix_5175_
inkcherry
inkcherry set clear_grad default to None, use foreach_zero
678a6e9e
inkcherry inkcherry requested a review from mrwyattii mrwyattii 1 year ago
inkcherry inkcherry requested a review from tjruwase tjruwase 1 year ago
inkcherry fix typo
6c568acf
inkcherry Merge branch 'master' into fix_5175_
d1ec5444
tjruwase
tjruwase commented on 2024-04-19
inkcherry fix condition&add detach
6b77946b
inkcherry Merge branch 'fix_5175_' of https://github.com/inkcherry/DeepSpeed in…
7c9c8a92
tjruwase
tjruwase commented on 2024-04-20
loadams Merge branch 'master' into fix_5175_
85f295dc
tjruwase
tjruwase approved these changes on 2024-04-22
loadams Merge branch 'master' into fix_5175_
32fc1ced
loadams Merge branch 'master' into fix_5175_
617e28aa
loadams loadams enabled auto-merge 1 year ago
loadams loadams merged c66bc426 into master 1 year ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
Milestone