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.

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

Fix ray hyperopt #1999

Merged
merged 22 commits into from
Jun 2, 2022
Merged

Fix ray hyperopt #1999

merged 22 commits into from
Jun 2, 2022

Conversation

ShreyaR
Copy link
Contributor

@ShreyaR ShreyaR commented May 5, 2022

No description provided.

@github-actions
Copy link

github-actions bot commented May 6, 2022

Unit Test Results

       6 files  ±0         6 suites  ±0   2h 22m 31s ⏱️ + 10m 34s
2 798 tests  - 1  2 766 ✔️ +2    32 💤  - 3  0 ±0 
8 394 runs   - 3  8 294 ✔️ +6  100 💤  - 9  0 ±0 

Results for commit 773b297. ± Comparison against base commit 698a0e0.

♻️ This comment has been updated with latest results.

@ShreyaR ShreyaR marked this pull request as ready for review May 12, 2022 18:22
@ShreyaR ShreyaR requested a review from tgaddair May 12, 2022 23:44
ludwig/api.py Outdated Show resolved Hide resolved
ludwig/api.py Outdated Show resolved Hide resolved
ludwig/backend/ray.py Outdated Show resolved Hide resolved
tests/integration_tests/test_hyperopt_ray_horovod.py Outdated Show resolved Hide resolved
ludwig/backend/ray.py Outdated Show resolved Hide resolved
res = ray.init(
num_cpus=4,
num_cpus=10,
Copy link
Collaborator

Choose a reason for hiding this comment

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

I don't think GH runners have 10 cpus. Do we really need this many? I think if it can't run with like 6 at most, there's probably a bug somewhere.

Copy link
Contributor Author

Choose a reason for hiding this comment

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

@tgaddair we would need at least one more CPU than the CPUs used in hyperopt trials (discussed in this slack thread).

We could artificially restrict hyperopt to run less than 6 trials so that there are CPUs available for dask computation by setting max_concurrent_trials to something less than 6. For now, I've changed the number of CPUs to 7.

ludwig/backend/ray.py Outdated Show resolved Hide resolved
ludwig/data/dataset/ray.py Outdated Show resolved Hide resolved
ludwig/backend/ray.py Outdated Show resolved Hide resolved
@ShreyaR ShreyaR force-pushed the fix-ray-hyperopt branch 2 times, most recently from c69ac82 to 002eb35 Compare May 26, 2022 00:26
Copy link
Collaborator

@tgaddair tgaddair left a comment

Choose a reason for hiding this comment

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

Looks good, only one small thing.

ludwig/data/dataframe/pandas.py Outdated Show resolved Hide resolved
@ShreyaR ShreyaR merged commit b59ce78 into master Jun 2, 2022
@ShreyaR ShreyaR deleted the fix-ray-hyperopt branch June 2, 2022 04:50
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants