mirror of
https://github.com/hpcaitech/ColossalAI.git
synced 2025-08-18 07:57:46 +00:00
[hotfix] fix norm type error in zero optimizer (#4795)
This commit is contained in:
parent
da15fdb9ca
commit
54b3ad8924
@ -221,8 +221,8 @@ def compute_norm(gradients: Tensor, dp_group: ProcessGroup, tp_group: ProcessGro
|
|||||||
else:
|
else:
|
||||||
total_norm = 0.0
|
total_norm = 0.0
|
||||||
for g in gradients:
|
for g in gradients:
|
||||||
param_norm = g.data.double().norm(2)
|
param_norm = g.data.double().norm(norm_type)
|
||||||
total_norm += param_norm.item() ** 2
|
total_norm += param_norm.item() ** norm_type
|
||||||
|
|
||||||
# Sum across all model parallel GPUs.
|
# Sum across all model parallel GPUs.
|
||||||
total_norm_cuda = torch.cuda.FloatTensor([float(total_norm)])
|
total_norm_cuda = torch.cuda.FloatTensor([float(total_norm)])
|
||||||
|
Loading…
Reference in New Issue
Block a user