[air] Deprecate some fields/classes that are supposed to be gone in 2…

….6. (ray-project#38794)

Signed-off-by: xwjiang2010 <xwjiang2010@gmail.com>
Signed-off-by: e428265 <arvind.chandramouli@lmco.com>

doc/source/train/getting-started-pytorch-lightning.rst

@@ -356,7 +356,7 @@ information about the training run, including the metrics and checkpoints report
 
     result.metrics     # The metrics reported during training.
     result.checkpoint  # The latest checkpoint reported during training.
-    result.log_dir     # The path where logs are stored.
+    result.path     # The path where logs are stored.
     result.error       # The exception that was raised, if training failed.
 
 .. TODO: Add results guide

doc/source/train/getting-started-pytorch.rst

@@ -278,7 +278,7 @@ information about the training run, including the metrics and checkpoints report
 
     result.metrics     # The metrics reported during training.
     result.checkpoint  # The latest checkpoint reported during training.
-    result.log_dir     # The path where logs are stored.
+    result.path     # The path where logs are stored.
     result.error       # The exception that was raised, if training failed.
 
 .. TODO: Add results guide

doc/source/train/getting-started-transformers.rst

@@ -280,7 +280,7 @@ information about the training run, including the metrics and checkpoints report
 
     result.metrics     # The metrics reported during training.
     result.checkpoint  # The latest checkpoint reported during training.
-    result.log_dir     # The path where logs are stored.
+    result.path     # The path where logs are stored.
     result.error       # The exception that was raised, if training failed.
 
 .. TODO: Add results guide

doc/source/tune/doc_code/key_concepts.py

@@ -148,7 +148,7 @@ def load_checkpoint(self, checkpoint_dir):
 
 best_result = results.get_best_result()  # Get best result object
 best_config = best_result.config  # Get best trial's hyperparameters
-best_logdir = best_result.log_dir  # Get best trial's logdir
+best_logdir = best_result.path  # Get best trial's result directory
 best_checkpoint = best_result.checkpoint  # Get best trial's best checkpoint
 best_metrics = best_result.metrics  # Get best trial's last results
 best_result_df = best_result.metrics_dataframe  # Get best result as pandas dataframe

doc/source/tune/examples/pbt_guide.ipynb

@@ -285,8 +285,8 @@
     "# Get the best trial result\n",
     "best_result = results_grid.get_best_result(metric=\"mean_accuracy\", mode=\"max\")\n",
     "\n",
-    "# Print `log_dir` where checkpoints are stored\n",
-    "print(\"Best result logdir:\", best_result.log_dir)\n",
+    "# Print `path` where checkpoints are stored\n",
+    "print('Best result path:', best_result.path)\n",
     "\n",
     "# Print the best trial `config` reported at the last iteration\n",
     "# NOTE: This config is just what the trial ended up with at the last iteration.\n",

python/ray/air/integrations/keras.py

@@ -5,7 +5,7 @@
 import ray
 from ray.train._internal.storage import _use_storage_context
 from ray.train.tensorflow import TensorflowCheckpoint, LegacyTensorflowCheckpoint
-from ray.util.annotations import PublicAPI, Deprecated
+from ray.util.annotations import PublicAPI
 
 
 class _Callback(KerasCallback):
@@ -183,51 +183,3 @@ def _get_reported_metrics(self, logs: Dict) -> Dict:
 
         assert isinstance(reported_metrics, dict)
         return reported_metrics
-
-
-@Deprecated
-class Callback(_Callback):
-    """
-    Keras callback for Ray AIR reporting and checkpointing.
-
-    You can use this in both TuneSession and TrainSession.
-
-    Example:
-        .. code-block: python
-
-            ############# Using it in TrainSession ###############
-            from ray.air.integrations.keras import Callback
-            def train_loop_per_worker():
-                strategy = tf.distribute.MultiWorkerMirroredStrategy()
-                with strategy.scope():
-                    model = build_model()
-                    #model.compile(...)
-                model.fit(dataset_shard, callbacks=[Callback()])
-
-    Args:
-        metrics: Metrics to report. If this is a list, each item describes
-            the metric key reported to Keras, and it will reported under the
-            same name. If this is a dict, each key will be the name reported
-            and the respective value will be the metric key reported to Keras.
-            If this is None, all Keras logs will be reported.
-        on: When to report metrics. Must be one of
-            the Keras event hooks (less the ``on_``), e.g.
-            "train_start", or "predict_end". Defaults to "epoch_end".
-        frequency: Checkpoint frequency. If this is an integer `n`,
-            checkpoints are saved every `n` times each hook was called. If
-            this is a list, it specifies the checkpoint frequencies for each
-            hook individually.
-
-    """
-
-    def __init__(
-        self,
-        metrics: Optional[Union[str, List[str], Dict[str, str]]] = None,
-        on: Union[str, List[str]] = "epoch_end",
-        frequency: Union[int, List[int]] = 1,
-    ):
-        # TODO: Remove this class in 2.6.
-        raise DeprecationWarning(
-            "`ray.air.integrations.keras.Callback` is deprecated. Use "
-            "`ray.air.integrations.keras.ReportCheckpointCallback` instead.",
-        )

python/ray/air/result.py

@@ -2,7 +2,6 @@
 import json
 import pandas as pd
 import pyarrow
-import warnings
 from dataclasses import dataclass
 from pathlib import Path
 from typing import TYPE_CHECKING, Any, Dict, List, Optional, Tuple, Union
@@ -13,7 +12,6 @@
     EXPR_PROGRESS_FILE,
     EXPR_ERROR_PICKLE_FILE,
 )
-from ray.util import log_once
 from ray.util.annotations import PublicAPI
 
 import logging
@@ -60,22 +58,6 @@ class Result:
     _remote_path: Optional[str] = None
     _storage_filesystem: Optional[pyarrow.fs.FileSystem] = None
     _items_to_repr = ["error", "metrics", "path", "filesystem", "checkpoint"]
-    # Deprecate: raise in 2.5, remove in 2.6
-    log_dir: Optional[Path] = None
-
-    def __post_init__(self):
-        if self.log_dir and log_once("result_log_dir_deprecated"):
-            warnings.warn(
-                "The `Result.log_dir` property is deprecated. "
-                "Use `local_path` instead."
-            )
-            self._local_path = str(self.log_dir)
-
-        # Duplicate for retrieval
-        self.log_dir = Path(self._local_path) if self._local_path else None
-        # Backwards compatibility: Make sure to cast Path to string
-        # Deprecate: Remove this line after 2.6
-        self._local_path = str(self._local_path) if self._local_path else None
 
     @property
     def config(self) -> Optional[Dict[str, Any]]:

python/ray/air/tests/test_keras_callback.py

@@ -7,7 +7,7 @@
 
 import ray
 from ray import train
-from ray.air.integrations.keras import Callback, ReportCheckpointCallback
+from ray.air.integrations.keras import ReportCheckpointCallback
 from ray.train.constants import TRAIN_DATASET_KEY
 from ray.train import ScalingConfig
 from ray.train.tensorflow import (
@@ -207,11 +207,6 @@ def test_keras_callback_e2e():
     predictor.predict(data=items)
 
 
-def test_keras_callback_is_deprecated():
-    with pytest.raises(DeprecationWarning):
-        Callback()
-
-
 if __name__ == "__main__":
     import sys
 

python/ray/train/examples/mlflow_simple_example.py

@@ -1,3 +1,4 @@
+from pathlib import Path
 from ray import train
 from ray.train import ScalingConfig, RunConfig
 from ray.train.torch import TorchTrainer
@@ -41,7 +42,7 @@ def train_func():
 
 # Print the latest run directory and keep note of it.
 # For example: /home/ubuntu/ray_results/TorchTrainer_2022-06-13_20-31-06
-print("Run directory:", result.log_dir.parent)  # TensorBoard is saved in parent dir
+print("Run directory:", Path(result.path).parent)  # TensorBoard is saved in parent dir
 
 # How to visualize the logs
 

python/ray/train/tensorflow/tensorflow_checkpoint.py

@@ -296,7 +296,6 @@ def from_saved_model(
     def get_model(
         self,
         model: Optional[Union[tf.keras.Model, Callable[[], tf.keras.Model]]] = None,
-        model_definition: Optional[Callable[[], tf.keras.Model]] = None,
     ) -> tf.keras.Model:
         """Retrieve the model stored in this checkpoint.
 
@@ -308,13 +307,6 @@ def get_model(
         Returns:
             The Tensorflow Keras model stored in the checkpoint.
         """
-        # TODO: Remove `model_definition` in 2.6.
-        if model_definition is not None:
-            raise DeprecationWarning(
-                "The `model_definition` parameter is deprecated. Use the `model` "
-                "parameter instead."
-            )
-
         if model is not None and self._flavor is not self.Flavor.MODEL_WEIGHTS:
             warnings.warn(
                 "TensorflowCheckpoint was created from "

python/ray/train/tests/test_lightning_trainer_restore.py

@@ -1,8 +1,8 @@
 import os
 import numpy as np
+from pathlib import Path
 import pytest
 import pytorch_lightning as pl
-from pathlib import Path
 from pytorch_lightning.callbacks import ModelCheckpoint
 
 import ray

python/ray/tune/tests/test_tuner.py

@@ -423,7 +423,7 @@ def train_func(config):
     results = tuner.fit()
     assert not results.errors
     for result in results:
-        artifact_data = open(result.log_dir / "write.txt", "r").read()
+        artifact_data = open(os.path.join(result.path, "write.txt"), "r").read()
         assert artifact_data == f"{result.config['id']}"
 
 
@@ -467,7 +467,7 @@ def train_func(config):
     results = tuner.fit()
     assert not results.errors
     for result in results:
-        artifact_data = open(result.log_dir / "write.txt", "r").read()
+        artifact_data = open(os.path.join(result.path, "write.txt"), "r").read()
         assert artifact_data == f"{result.config['id']}"
 
 

python/ray/tune/tests/test_tuner_restore.py

@@ -915,6 +915,7 @@ def failing_fn(config):
             if path.startswith("experiment_state")
         ]
     )
+
     assert num_experiment_checkpoints == 2
 
     num_trial_checkpoints = len(

python/ray/tune/tests/tutorial.py

@@ -133,7 +133,7 @@ def train_mnist(config):
 # __eval_func_end__
 
 # __plot_begin__
-dfs = {result.log_dir: result.metrics_dataframe for result in results}
+dfs = {result.path: result.metrics_dataframe for result in results}
 [d.mean_accuracy.plot() for d in dfs.values()]
 # __plot_end__
 
@@ -149,7 +149,7 @@ def train_mnist(config):
 results = tuner.fit()
 
 # Obtain a trial dataframe from all run trials of this `tune.run` call.
-dfs = {result.log_dir: result.metrics_dataframe for result in results}
+dfs = {result.path: result.metrics_dataframe for result in results}
 # __run_scheduler_end__
 
 # fmt: off
@@ -190,7 +190,7 @@ def train_mnist(config):
 # __run_analysis_begin__
 import os
 
-logdir = results.get_best_result("mean_accuracy", mode="max").log_dir
+logdir = results.get_best_result("mean_accuracy", mode="max").path
 state_dict = torch.load(os.path.join(logdir, "model.pth"))
 
 model = ConvNet()