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

Do not reset CUDA context after UCX tests #8201

Merged
merged 4 commits into from
Sep 25, 2023
Merged

Do not reset CUDA context after UCX tests #8201

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
40285e7
Do not reset CUDA context after UCX tests
pentschev Sep 21, 2023
89aa552
Prevent cuDF `FutureWarning` on empty series datatype
pentschev Sep 21, 2023
0f1c4da
Downgrade to aws-sdk-cpp<1.11
pentschev Sep 22, 2023
119dc14
Revert "Downgrade to aws-sdk-cpp<1.11"
pentschev Sep 23, 2023
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
2 changes: 2 additions & 0 deletions continuous_integration/gpuci/build.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,8 @@ gpuci_logger "Activate conda env"
. /opt/conda/etc/profile.d/conda.sh
conda activate dask

mamba install -y 'aws-sdk-cpp<1.11'
Copy link
Member

Choose a reason for hiding this comment

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

Is this due to some upstream issue? The actual change here looks totally fine, but it might be worth a comment with extra context

Copy link
Member Author

Choose a reason for hiding this comment

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

Indeed, this is due to aws/aws-sdk-cpp#2681, but the change here was just to test/confirm this. For now we will likely resolve the issue via rapidsai/cudf#14173 and revert this change here.

Sorry for not commenting here earlier, I was trying to get this in before bringing people in to review the changes here. 🙂

Copy link
Member

Choose a reason for hiding this comment

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

No worries. I just noticed that rapidsai/cudf#14173 is merged -- does that mean we no longer need this here?

Copy link
Member Author

Choose a reason for hiding this comment

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

It does now, we first had to fix some issues (mark old cuDF packages as broken and build new gpuCI docker images). I reverted the aws-sdk-cpp changes now, we just need to wait and see if everything passes and we should be good to merge this. 🙂


gpuci_logger "Install distributed"
python -m pip install -e .

Expand Down
2 changes: 1 addition & 1 deletion distributed/comm/tests/test_ucx.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -179,7 +179,7 @@ async def test_ucx_deserialize(ucx_loop):
"g",
[
lambda cudf: cudf.Series([1, 2, 3]),
lambda cudf: cudf.Series([]),
lambda cudf: cudf.Series([], dtype=object),
lambda cudf: cudf.DataFrame([]),
lambda cudf: cudf.DataFrame([1]).head(0),
lambda cudf: cudf.DataFrame([1.0]).head(0),
Expand Down
11 changes: 0 additions & 11 deletions distributed/utils_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2156,17 +2156,6 @@ def ucx_loop():
import distributed.comm.ucx

distributed.comm.ucx.ucp = None
# If the test created a context, clean it up.
# TODO: should we check if there's already a context _before_ the test runs?
# I think that would be useful.
from distributed.diagnostics.nvml import has_cuda_context

ctx = has_cuda_context()
if ctx.has_context:
import numba.cuda

ctx = numba.cuda.current_context()
ctx.device.reset()


def wait_for_log_line(
Expand Down