Skip to content

bugfix, add skipIf for multigpu tests #62499

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
LiYuRio merged 2 commits into from
May 16, 2024
Merged

bugfix, add skipIf for multigpu tests #62499

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3f89fc5
bugfix, add skipIf for multigpu tests
jeng1220 Jan 25, 2024
2378025
Bugfix test_op with single GPU
jeng1220 Feb 16, 2024
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 test/collective/test_communication_api_base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,9 +22,14 @@
import tempfile
import unittest

import paddle


class CommunicationTestDistBase(unittest.TestCase):
def setUp(self, save_log_dir=None, num_of_devices=2, timeout=120, nnode=1):
if num_of_devices > paddle.device.cuda.device_count():
self.skipTest("number of GPUs is not enough")
Comment on lines +30 to +31
Copy link
Contributor

@LiYuRio LiYuRio Mar 7, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think need this if-condition here because all the test-case inherited this base class only use 2-cards. And CI machines which run these test-cases only has two devices either.

Copy link
Collaborator Author

@jeng1220 jeng1220 Mar 8, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@LiYuRio ,

Thanks for the response.

CI machines which run these test-cases only has two devices either.

Inside Baidu, it is true but for other developers, it might not be true. This line can give other developers a simple, more meaningful reason to show why unit tests cannot run, instead of a long bug message. Does it make sense?

Copy link
Contributor

@LiYuRio LiYuRio Mar 13, 2024 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

ok, but if we add skipTest here, is it necessary to add skipTest to all distributed test case? Not only in communication test case.

Copy link
Collaborator Author

@jeng1220 jeng1220 Apr 25, 2024

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@LiYuRio ,
Yes, it will be best. I just found another test (test/legacy_test/op_test.py) has the same issue. This is all I have found for now.


self._python_interp = sys.executable
self._save_log_dir = save_log_dir
self._log_dir = tempfile.TemporaryDirectory()
Expand Down
7 changes: 7 additions & 0 deletions test/legacy_test/op_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3199,6 +3199,13 @@ def check_grad_with_place(
python_api_info=python_api_info,
)
runtime_envs = get_subprocess_runtime_envs(place)

num_devices = len(
runtime_envs["CUDA_VISIBLE_DEVICES"].split(",")
)
if num_devices > paddle.device.cuda.device_count():
self.skipTest("number of GPUs is not enough")

start_command = get_subprocess_command(
runtime_envs["CUDA_VISIBLE_DEVICES"],
generated_grad_test_path,
Expand Down