DeepSpeed
re-introduce: stage3: efficient compute of scaled_global_grad_norm
#5493
Merged

re-introduce: stage3: efficient compute of scaled_global_grad_norm #5493

lekurile merged 2 commits into deepspeedai:master from nelyahu:offload_fix
nelyahu
Revert "Revert "stage3: efficient compute of scaled_global_grad_norm …
63a89be1
fix for complete_grad_norm_calc in stage3
95aee34f
nelyahu nelyahu requested a review from tjruwase tjruwase 1 year ago
nelyahu nelyahu requested a review from mrwyattii mrwyattii 1 year ago
nelyahu
lekurile
nelyahu
lekurile
lekurile lekurile requested a review from lekurile lekurile 1 year ago
lekurile
lekurile approved these changes on 2024-05-02
lekurile lekurile merged 90793aab into master 1 year ago
nelyahu nelyahu deleted the offload_fix branch 1 year ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone