[train][release] Add training ingest soak test #57120
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Timothy Seah <tseah@anyscale.com>
TimothySeah
force-pushed
the
tseah/train-soak-test
branch
from
d7e6e97 to
3c376f8
Compare
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Code Review
This pull request adds a new soak test for training ingest. The test runs an image classification task for 1000 epochs on a 16-GPU cluster. My review focuses on the configuration of this new test. I've suggested removing a debug flag to prevent excessive logging and recommended adjusting the test timeout to be more conservative to avoid tying up resources unnecessarily. Overall, the changes are straightforward and look good with these minor adjustments.
Signed-off-by: Timothy Seah <tseah@anyscale.com>
Signed-off-by: Timothy Seah <tseah@anyscale.com>
Signed-off-by: Timothy Seah <tseah@anyscale.com>
Signed-off-by: Timothy Seah <tseah@anyscale.com>
Signed-off-by: Timothy Seah <tseah@anyscale.com>
Signed-off-by: Timothy Seah <tseah@anyscale.com>
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
This comment was marked as outdated.
This comment was marked as outdated.
Sorry, something went wrong.
justinvyu
reviewed
Oct 14, 2025
Comment on lines
+18
to
+22
| - name: worker_node_cpu | ||
| instance_type: m5.4xlarge | ||
| max_workers: 4 | ||
| min_workers: 4 | ||
| use_spot: false |
Contributor
There was a problem hiding this comment.
do we need to add cpu nodes for this test?
Contributor
Author
There was a problem hiding this comment.
See the testing notes (#57120 (comment)) - doing this kept memory growth low. Lmk if this is fine.
release/train_tests/benchmark/compute_configs/compute_gpu_4x4_cpu_4_aws.yaml
Show resolved
Hide resolved
Signed-off-by: Timothy Seah <tseah@anyscale.com>
justinvyu
approved these changes
Oct 15, 2025
Comment on lines
+18
to
+22
| - name: worker_node_cpu | ||
| instance_type: m5.4xlarge | ||
| max_workers: 4 | ||
| min_workers: 4 | ||
| use_spot: false |
justinyeh1995
pushed a commit
to justinyeh1995/ray
that referenced
this pull request
Oct 20, 2025
Run `full_training.parquet` for 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold. --------- Signed-off-by: Timothy Seah <tseah@anyscale.com>
xinyuangui2
pushed a commit
to xinyuangui2/ray
that referenced
this pull request
Oct 22, 2025
Run `full_training.parquet` for 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold. --------- Signed-off-by: Timothy Seah <tseah@anyscale.com> Signed-off-by: xgui <xgui@anyscale.com>
elliot-barn
pushed a commit
that referenced
this pull request
Oct 23, 2025
Run `full_training.parquet` for 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold. --------- Signed-off-by: Timothy Seah <tseah@anyscale.com> Signed-off-by: elliot-barn <elliot.barnwell@anyscale.com>
landscapepainter
pushed a commit
to landscapepainter/ray
that referenced
this pull request
Nov 17, 2025
Run `full_training.parquet` for 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold. --------- Signed-off-by: Timothy Seah <tseah@anyscale.com>
Aydin-ab
pushed a commit
to Aydin-ab/ray-aydin
that referenced
this pull request
Nov 19, 2025
Run `full_training.parquet` for 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold. --------- Signed-off-by: Timothy Seah <tseah@anyscale.com> Signed-off-by: Aydin Abiar <aydin@anyscale.com>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
Run
full_training.parquetfor 50 epochs as a soak test. In the future, we can make this more meaningful by tracking metrics (e.g. memory usage) and failing if they exceed a certain threshold.Testing
OOD run: https://buildkite.com/ray-project/release/builds/61277#0199a61f-b2aa-4782-8eeb-83429cc50a36.
head node cpu 0 run: https://console.anyscale-staging.com/cld_kvedZWag2qA8i5BjxUevf5i7/prj_92c7b71w55flm6gv6imv4m6vqg/jobs/prodjob_m5r3bthpdpf4x3gfpn4fcmyxvh/data?job-logs-section-tabs=application_logs&job-tab=metrics, https://buildkite.com/ray-project/release/builds/62401#_. No more disk growth but there is some memory growth that tapers off and fortunately doesn't OOM the test
head node cpu 0, gpu worker cpu 0, cpu worker run: https://console.anyscale-staging.com/cld_kvedZWag2qA8i5BjxUevf5i7/prj_92c7b71w55flm6gv6imv4m6vqg/jobs/prodjob_1akz6zmrr3mtgf1cpqyhc7mwgm?job-logs-section-tabs=application_logs&job-tab=metrics, https://buildkite.com/ray-project/release/builds/62479#0199c5cf-56fa-44c2-9895-3b8a863fe33e. Still no disk growth, and memory growth also stops much earlier
Note
Adds a weekly GPU soak test that runs the image_classification parquet ingest benchmark for 1000 epochs with Ray Data.
training_ingest_benchmark-soak-testinrelease/release_tests.yaml:ml; GPU clustercompute_configs/compute_gpu_4x4_aws.yaml.RAY_DATA_DEBUG_RESOURCE_MANAGER=1.train_tests/benchmark/train_benchmark.pywith--task=image_classification --dataloader_type=ray_data --image_classification_data_format=parquet --num_workers=16 --num_epochs=1000;timeout: 86400.Written by Cursor Bugbot for commit d7e6e97. This will update automatically on new commits. Configure here.