Skip to content

fix(npu): synchronize stream after backward for dp replicate training #60

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
danieldale2026 wants to merge 1 commit into
base: feat/npu
Choose a base branch
from
+3 −0
Open

fix(npu): synchronize stream after backward for dp replicate training #60

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions mova/engine/trainer/accelerate/accelerate_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@

import os, re
import torch
import torch_npu
from tqdm import tqdm

try:
Expand Down Expand Up @@ -411,6 +412,8 @@ def train(self):
loss = loss_dict["loss"]

self.accelerator.backward(loss)
# Synchronize NPU stream to avoid async ordering issues with DP replicate training.
torch_npu.npu.current_stream().synchronize()

if self.gradient_clip_norm > 0:
if self.accelerator.sync_gradients:
Expand Down