github
/
ColossalAI
mirror of https://github.com/hpcaitech/ColossalAI
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

[hotfix] fix norm type error in zero optimizer (#4795)

pull/4815/head
littsk 2023-09-27 10:35:24 +08:00 committed by GitHub
parent da15fdb9ca
commit 54b3ad8924
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
colossalai/zero/low_level/_utils.py
View File

@ -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)])