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

Jump to bottom

[RAY AIR] set the correct gpu id in TorchTrainer #26493

Merged
merged 15 commits into from
Jul 19, 2022
Merged

[RAY AIR] set the correct gpu id in TorchTrainer #26493

Show file tree
Hide file tree
Changes from 7 commits
Commits
Show all changes
15 commits
Select commit Hold shift + click to select a range
72108b7
use local rank for cuda device
JiahaoYao Jul 13, 2022
ee70c88
adding the test
JiahaoYao Jul 13, 2022
e72378c
lint
JiahaoYao Jul 13, 2022
6a8e934
fix ci
JiahaoYao Jul 13, 2022
e4e967e
revert the test
JiahaoYao Jul 13, 2022
4a8112e
gpu test
JiahaoYao Jul 13, 2022
e2911c7
not use private utils
JiahaoYao Jul 13, 2022
09f2fa5
adding the test and comment
JiahaoYao Jul 18, 2022
0f9d437
Merge remote-tracking branch 'upstream/master' into 0712_gpuid
JiahaoYao Jul 18, 2022
de881c4
fix error
JiahaoYao Jul 18, 2022
0727edd
adding the tune test
JiahaoYao Jul 18, 2022
8975238
monkeypatch
JiahaoYao Jul 18, 2022
ab129db
update
richardliaw Jul 18, 2022
5d680e4
update
richardliaw Jul 18, 2022
d4263ae
epoch == 2
JiahaoYao Jul 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
4 changes: 4 additions & 0 deletions python/ray/train/tests/test_gpu.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -345,6 +345,10 @@ def test_tune_fashion_mnist_gpu(ray_start_4_cpus_2_gpus):
torch_fashion_mnist(num_workers=2, use_gpu=True, num_samples=1)


def test_concurrent_tune_fashion_mnist_gpu(ray_start_4_cpus_2_gpus):
torch_fashion_mnist(num_workers=1, use_gpu=True, num_samples=2)


def test_tune_tensorflow_mnist_gpu(ray_start_4_cpus_2_gpus):
tune_tensorflow_mnist(num_workers=2, use_gpu=True, num_samples=1)

Expand Down
9 changes: 8 additions & 1 deletion python/ray/train/torch/train_loop_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -466,8 +466,15 @@ def get_device(self) -> torch.device:
"""Gets the correct torch device to use for training."""
if torch.cuda.is_available():
gpu_ids = ray.get_gpu_ids()
richardliaw marked this conversation as resolved.
Show resolved Hide resolved

if len(gpu_ids) > 0:
device_id = gpu_ids[0]
gpu_id = gpu_ids[0]
cuda_visible_str = os.environ.get("CUDA_VISIBLE_DEVICES", None)
assert (
cuda_visible_str and cuda_visible_str != "NoDevFiles"
), "CUDA_VISIBLE_DEVICES is not set or empty."
cuda_visible_list = list(map(int, cuda_visible_str.split(",")))
device_id = cuda_visible_list.index(gpu_id)
Copy link
Contributor

Choose a reason for hiding this comment

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

Will this fail if cuda_visible_list is empty?

Copy link
Contributor Author

Choose a reason for hiding this comment

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

Thanks @amogkam for the review, about to modify the code accordingly.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

len(gpu_ids) > 0 will ensure this?

else:
# If called on the driver or outside of Ray Train, return the
# 0th device.
Expand Down