mirror of
https://github.com/zebrajr/pytorch.git
synced 2025-12-06 12:20:52 +01:00
644bc69530
Right now DCP API requires users to create StorageWriter and StorageReader for every API call. This PR allows users to only pass the checkpointer_id (a path) and use it to read/write a checkpoint without creating a StorageReader and Writer. Differential Revision: [D52740556](https://our.internmc.facebook.com/intern/diff/D52740556/) Pull Request resolved: https://github.com/pytorch/pytorch/pull/117772 Approved by: https://github.com/wz337 ghstack dependencies: #116248
105 lines
2.7 KiB
Python
105 lines
2.7 KiB
Python
# Owner(s): ["oncall: distributed"]
|
|
|
|
|
|
import os
|
|
import shutil
|
|
|
|
import torch
|
|
import torch.distributed as dist
|
|
import torch.distributed.checkpoint as dcp
|
|
import torch.multiprocessing as mp
|
|
import torch.nn as nn
|
|
from torch.distributed._tensor.device_mesh import init_device_mesh
|
|
from torch.distributed.checkpoint.state_dict import (
|
|
_patch_model_state_dict,
|
|
_patch_optimizer_state_dict,
|
|
)
|
|
from torch.distributed.fsdp import FullyShardedDataParallel as FSDP
|
|
|
|
|
|
CHECKPOINT_DIR = f"~/{os.environ['LOGNAME']}/checkpoint"
|
|
|
|
|
|
class Model(torch.nn.Module):
|
|
def __init__(self):
|
|
super().__init__()
|
|
torch.manual_seed(0)
|
|
self.net1 = nn.Sequential(nn.Linear(8, 16), nn.ReLU())
|
|
self.net2 = nn.Sequential(nn.Linear(16, 32), nn.ReLU())
|
|
self.net3 = nn.Linear(32, 64)
|
|
self.net4 = nn.Sequential(nn.ReLU(), nn.Linear(64, 8))
|
|
|
|
def forward(self, x):
|
|
return self.net4(self.net3(self.net2(self.net1(x))))
|
|
|
|
def get_input(self):
|
|
return torch.rand(8, 8, device="cuda")
|
|
|
|
|
|
def _make_stateful(model, optim):
|
|
_patch_model_state_dict(model)
|
|
_patch_optimizer_state_dict(model, optimizers=optim)
|
|
|
|
|
|
def _train(model, optim, train_steps=1):
|
|
torch.manual_seed(0)
|
|
loss = None
|
|
for _ in range(train_steps):
|
|
loss = model(model.get_input()).sum()
|
|
loss.backward()
|
|
optim.step()
|
|
optim.zero_grad()
|
|
|
|
return loss
|
|
|
|
|
|
def _init_model(device, world_size):
|
|
device_mesh = init_device_mesh(device, (world_size,))
|
|
model = Model().cuda()
|
|
model = FSDP(
|
|
model,
|
|
device_mesh=device_mesh,
|
|
use_orig_params=True,
|
|
)
|
|
optim = torch.optim.Adam(model.parameters(), lr=0.1)
|
|
_make_stateful(model, optim)
|
|
|
|
return model, optim
|
|
|
|
|
|
def run(rank, world_size, device="cuda"):
|
|
# Set up world pg
|
|
os.environ["MASTER_ADDR"] = "localhost"
|
|
os.environ["MASTER_PORT"] = "12355"
|
|
|
|
dist.init_process_group("cpu:gloo,cuda:nccl", rank=rank, world_size=world_size)
|
|
torch.cuda.set_device(rank)
|
|
|
|
model, optim = _init_model(device, world_size)
|
|
_train(model, optim, train_steps=2)
|
|
|
|
dcp.save(
|
|
state_dict={"model": model, "optimizer": optim},
|
|
checkpoint_id=CHECKPOINT_DIR,
|
|
)
|
|
|
|
# presumably do something else
|
|
model, optim = _init_model(device, world_size)
|
|
dcp.load(
|
|
state_dict={"model": model, "optimizer": optim},
|
|
checkpoint_id=CHECKPOINT_DIR,
|
|
)
|
|
_train(model, optim, train_steps=2)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
world_size = torch.cuda.device_count()
|
|
print(f"Running stateful checkpoint example on {world_size} devices.")
|
|
shutil.rmtree(CHECKPOINT_DIR, ignore_errors=True)
|
|
mp.spawn(
|
|
run,
|
|
args=(world_size,),
|
|
nprocs=world_size,
|
|
join=True,
|
|
)
|