Skip to content

feat(examples): better torch distributed training example launch via mpi #77

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

Merged
jingxu9x merged 5 commits into from
Aug 7, 2025
Merged

feat(examples): better torch distributed training example launch via mpi #77

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
f6e1247
feat: direct get init_process_group paramters from ompi and environment
jingxu9x Aug 7, 2025
b815a4a
fix: pass lint
jingxu9x Aug 7, 2025
854f48a
fix: print avg loss
jingxu9x Aug 7, 2025
c4e4771
fix: only rank 0 print the progress status
jingxu9x Aug 7, 2025
c684e5d
fix: only rank 0 print the progress status
jingxu9x Aug 7, 2025
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
23 changes: 15 additions & 8 deletions advanced/pytorch-example/main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@
import torch.nn.functional as F
import torch.distributed as dist
import os
from torchvision import datasets, transforms
from torchvision import transforms
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.utils.data import DataLoader, DistributedSampler

Expand Down Expand Up @@ -36,13 +36,19 @@ def forward(self, x):
return F.log_softmax(x, dim=1)

def train():
master_addr = os.environ.get("MASTER_ADDR", "localhost")
master_port = os.environ.get("MASTER_PORT", "29500")
world_size = int(os.environ["OMPI_COMM_WORLD_SIZE"])
rank = int(os.environ["OMPI_COMM_WORLD_RANK"])
local_rank = rank % torch.cuda.device_count()

# Initialize process group
dist.init_process_group(backend="nccl")

# Get local rank from environment variable
local_rank = int(os.environ["LOCAL_RANK"])
rank = int(os.environ["RANK"])
world_size = int(os.environ["WORLD_SIZE"])
dist.init_process_group(
backend="nccl",
init_method=f"tcp://{master_addr}:{master_port}",
world_size=world_size,
rank=rank
)

# Set device
torch.cuda.set_device(local_rank)
Expand Down Expand Up @@ -77,7 +83,8 @@ def transform(example):
loss.backward()
optimizer.step()

if batch_idx % 10 == 0:
dist.all_reduce(loss, op=dist.ReduceOp.AVG)
if rank == 0 and batch_idx % 10 == 0:
print(f"Train Epoch: {epoch} [{batch_idx * len(data)}/{len(train_loader.dataset)} ({100. * batch_idx / len(train_loader):.0f}%)]\tLoss: {loss.item():.6f}")

if rank == 0:
Expand Down
Loading