Skip to content
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

Re-enable Lite CLI on Windows + PyTorch 1.13 #15645

Merged
merged 4 commits into from
Dec 19, 2022
Merged

Re-enable Lite CLI on Windows + PyTorch 1.13 #15645

Show file tree
Hide file tree
Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
4efbd6c
Re-enable Lite CLI on Windows + PyTorch 1.13
awaelchli Nov 11, 2022
18996cd
[pre-commit.ci] auto fixes from pre-commit.com hooks
pre-commit-ci[bot] Nov 11, 2022
528d2ba
Merge branch 'master' into bugfix/lite-cli-windows-1.13.1
awaelchli Dec 17, 2022
7cd87a7
Merge branch 'master' into bugfix/lite-cli-windows-1.13.1
justusschock Dec 19, 2022
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
10 changes: 0 additions & 10 deletions src/lightning_lite/cli.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,6 @@

from lightning_lite.accelerators import CPUAccelerator, CUDAAccelerator, MPSAccelerator
from lightning_lite.utilities.device_parser import _parse_gpu_ids
from lightning_lite.utilities.imports import _IS_WINDOWS, _TORCH_GREATER_EQUAL_1_13

_log = logging.getLogger(__name__)

Expand Down Expand Up @@ -148,15 +147,6 @@ def _get_num_processes(accelerator: str, devices: str) -> int:

def _torchrun_launch(args: Namespace, script_args: List[str]) -> None:
"""This will invoke `torchrun` programmatically to launch the given script in new processes."""

if _IS_WINDOWS and _TORCH_GREATER_EQUAL_1_13: # pragma: no cover
# TODO: remove once import issue is resolved: https://github.com/pytorch/pytorch/issues/85427
_log.error(
"On the Windows platform, this launcher is currently only supported on torch < 1.13 due to a bug"
" upstream: https://github.com/pytorch/pytorch/issues/85427"
)
raise SystemExit(1)

import torch.distributed.run as torchrun

if args.strategy == "dp":
Expand Down
22 changes: 1 addition & 21 deletions tests/tests_lite/test_cli.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,21 +16,10 @@
from unittest.mock import Mock

import pytest
import torch.distributed.run
from tests_lite.helpers.runif import RunIf

from lightning_lite.cli import _run_model
from lightning_lite.utilities.imports import _IS_WINDOWS, _TORCH_GREATER_EQUAL_1_13

if not (_IS_WINDOWS and _TORCH_GREATER_EQUAL_1_13):
import torch.distributed.run


def skip_windows_pt_1_13():
# https://github.com/pytorch/pytorch/issues/85427
return pytest.mark.skipif(
condition=(_IS_WINDOWS and _TORCH_GREATER_EQUAL_1_13),
reason="Torchelastic import bug in 1.13 affecting Windows",
)


@pytest.fixture
Expand All @@ -40,7 +29,6 @@ def fake_script(tmp_path):
return str(script)


@skip_windows_pt_1_13()
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
def test_cli_env_vars_defaults(monkeypatch, fake_script):
monkeypatch.setattr(torch.distributed, "run", Mock())
Expand All @@ -55,7 +43,6 @@ def test_cli_env_vars_defaults(monkeypatch, fake_script):
assert os.environ["LT_PRECISION"] == "32"


@skip_windows_pt_1_13()
@pytest.mark.parametrize("accelerator", ["cpu", "gpu", "cuda", pytest.param("mps", marks=RunIf(mps=True))])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
@mock.patch("lightning_lite.accelerators.cuda.num_cuda_devices", return_value=2)
Expand All @@ -67,7 +54,6 @@ def test_cli_env_vars_accelerator(_, accelerator, monkeypatch, fake_script):
assert os.environ["LT_ACCELERATOR"] == accelerator


@skip_windows_pt_1_13()
@pytest.mark.parametrize("strategy", ["dp", "ddp", "deepspeed"])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
@mock.patch("lightning_lite.accelerators.cuda.num_cuda_devices", return_value=2)
Expand All @@ -79,7 +65,6 @@ def test_cli_env_vars_strategy(_, strategy, monkeypatch, fake_script):
assert os.environ["LT_STRATEGY"] == strategy


@skip_windows_pt_1_13()
@pytest.mark.parametrize("devices", ["1", "2", "0,", "1,0", "-1"])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
@mock.patch("lightning_lite.accelerators.cuda.num_cuda_devices", return_value=2)
Expand All @@ -92,7 +77,6 @@ def test_cli_env_vars_devices_cuda(_, devices, monkeypatch, fake_script):


@RunIf(mps=True)
@skip_windows_pt_1_13()
@pytest.mark.parametrize("accelerator", ["mps", "gpu"])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
def test_cli_env_vars_devices_mps(accelerator, monkeypatch, fake_script):
Expand All @@ -103,7 +87,6 @@ def test_cli_env_vars_devices_mps(accelerator, monkeypatch, fake_script):
assert os.environ["LT_DEVICES"] == "1"


@skip_windows_pt_1_13()
@pytest.mark.parametrize("num_nodes", ["1", "2", "3"])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
def test_cli_env_vars_num_nodes(num_nodes, monkeypatch, fake_script):
Expand All @@ -114,7 +97,6 @@ def test_cli_env_vars_num_nodes(num_nodes, monkeypatch, fake_script):
assert os.environ["LT_NUM_NODES"] == num_nodes


@skip_windows_pt_1_13()
@pytest.mark.parametrize("precision", ["64", "32", "16", "bf16"])
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
def test_cli_env_vars_precision(precision, monkeypatch, fake_script):
Expand All @@ -125,7 +107,6 @@ def test_cli_env_vars_precision(precision, monkeypatch, fake_script):
assert os.environ["LT_PRECISION"] == precision


@skip_windows_pt_1_13()
@mock.patch.dict(os.environ, os.environ.copy(), clear=True)
def test_cli_torchrun_defaults(monkeypatch, fake_script):
torchrun_mock = Mock()
Expand All @@ -145,7 +126,6 @@ def test_cli_torchrun_defaults(monkeypatch, fake_script):
)


@skip_windows_pt_1_13()
@pytest.mark.parametrize(
"devices,expected",
[
Expand Down