[tune] Storage: 🐙 🧠 Tune tests and examples {using RLlib} migration (ray-project#38895)

Signed-off-by: Kai Fricke <kai@anyscale.com>
Co-authored-by: matthewdeng <matt@anyscale.com>

Author: 2 people

.buildkite/pipeline.ml.yml

@@ -341,7 +341,6 @@
     - TUNE_TESTING=1 ./ci/env/install-dependencies.sh
     - ./ci/env/env_info.sh
     - bazel test --config=ci $(./ci/run/bazel_export_options) --build_tests_only --test_tag_filters=-gpu_only,rllib
-      --test_env=RAY_AIR_NEW_PERSISTENCE_MODE=0
       python/ray/tune/...
 
 - label: ":octopus: ML library integrations tests and examples."

python/ray/tune/analysis/experiment_analysis.py

@@ -680,7 +680,11 @@ def _retrieve_rows(
         assert not mode or metric
         rows = {}
         for path, df in self.trial_dataframes.items():
-            if mode == "max":
+            if df.empty:
+                continue
+            if metric not in df:
+                idx = -1
+            elif mode == "max":
                 idx = df[metric].idxmax()
             elif mode == "min":
                 idx = df[metric].idxmin()

python/ray/tune/experiment/trial.py

@@ -1082,7 +1082,7 @@ def should_checkpoint(self):
 
     def has_checkpoint(self):
         if _use_storage_context():
-            return self.checkpoint.path is not None
+            return self.checkpoint is not None
         return self.checkpoint.dir_or_data is not None
 
     def clear_checkpoint(self):