Skip to content

[fix] ShardedDDP - cpu testfix - remove Gloo/CPU #350

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
blefaudeux merged 4 commits into from
Feb 2, 2021
+12 −6
Merged

[fix] ShardedDDP - cpu testfix - remove Gloo/CPU #350

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
24713da
removing the cpu test, not too interesting anyway
blefaudeux Feb 1, 2021
2dabbcc
removing gloo for now
blefaudeux Feb 1, 2021
f37fef2
review, thanks Min
blefaudeux Feb 2, 2021
08beb60
now with a reason
blefaudeux Feb 2, 2021
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
5 changes: 5 additions & 0 deletions fairscale/utils/testing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@
import multiprocessing
import os
import random
import sys
import tempfile
from typing import Any, Callable, Dict, List, Optional, Tuple

Expand All @@ -53,6 +54,10 @@
not torch.cuda.is_available() or torch.cuda.device_count() < 2, reason="multiple GPUs required"
)

skip_if_py38 = pytest.mark.skipif(
sys.version_info.major == 3 and sys.version_info.minor == 8, reason="Python3.8 is skipped"
)

_, filename_mpi = tempfile.mkstemp()


Expand Down
13 changes: 7 additions & 6 deletions tests/nn/data_parallel/test_sharded_ddp.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,7 +21,7 @@

from fairscale.nn.data_parallel import ShardedDataParallel
from fairscale.optim import OSS
from fairscale.utils.testing import GPT2, skip_if_no_cuda, skip_if_single_gpu
from fairscale.utils.testing import GPT2, skip_if_no_cuda, skip_if_py38, skip_if_single_gpu


def run_one_step(rank, world_size, backend, device, temp_file_name):
Expand Down Expand Up @@ -112,16 +112,17 @@ def run_test(backend, device, world_size=2):
mp.spawn(run_one_step, args=(world_size, backend, device, temp_file_name), nprocs=world_size, join=True)


def test_step_on_cpu():
run_test(backend=dist.Backend.GLOO, device=torch.device("cpu"), world_size=4)


@skip_if_no_cuda
@skip_if_single_gpu
def test_step_on_gpu():
def test_step_gpu():
run_test(backend=dist.Backend.NCCL, device=torch.device("cuda"))


@skip_if_py38
def test_step_cpu():
run_test(backend=dist.Backend.GLOO, device=torch.device("cpu"))


def run_ddp_parity(rank, world_size, backend, temp_file_name):
url = "file://" + temp_file_name
dist.init_process_group(init_method=url, backend=backend, rank=rank, world_size=world_size)
Expand Down