Replace ray.state.nodes() with ray.nodes().

[justinxzhao]

No description provided.

[github-actions]

Unit Test Results

       6 files  ±0         6 suites  ±0   2h 28m 34s ⏱️ + 10m 31s
2 832 tests +1  2 798 ✔️ +2    34 💤 ±0  0 ❌  - 1 
8 496 runs  +3  8 390 ✔️ +4  106 💤 ±0  0 ❌  - 1 

Results for commit e087af3. ± Comparison against base commit 18f2c2f.

♻️ This comment has been updated with latest results.

[tgaddair]

https://github.com/ludwig-ai/ludwig/runs/6998208225?check_suite_focus=true#step:10:2220

ValueError: Metadata mismatch found in `from_delayed`.

Maybe @clarkzinzow knows why this is suddenly popping up?

[justinxzhao]

@tgaddair @clarkzinzow There's some flakiness with the hyperopt integration tests.

After re-running the CI, the test with the Metadata mismatch error went away.

However, we're still running into an error with tests/integration_tests/test_hyperopt_ray.py::test_hyperopt_ray_mlflow

Here's the error

>   from ray.util.ml_utils.mlflow import MLflowLoggerUtil as _MLflowLoggerUtil
E   ImportError: cannot import name 'MLflowLoggerUtil' from 'ray.util.ml_utils.mlflow' (/opt/hostedtoolcache/Python/3.9.13/x64/lib/python3.9/site-packages/ray/util/ml_utils/mlflow.py)

[justinxzhao]

The MLflowLoggerUtil issue was due to a temporary workaround that was implemented in Ludwig that is no longer needed with newer versions of Ray.