adapt to kt 1.0.3 (#1593)

Co-authored-by: Haifeng Jin <[email protected]>

autokeras/auto_model.py

@@ -111,7 +111,7 @@ class AutoModel(object):
         seed: Int. Random seed.
         max_model_size: Int. Maximum number of scalars in the parameters of a
             model. Models larger than this are rejected.
-        **kwargs: Any arguments supported by kerastuner.Tuner.
+        **kwargs: Any arguments supported by keras_tuner.Tuner.
     """
 
     def __init__(

autokeras/blocks/basic.py

@@ -16,7 +16,7 @@
 from typing import Union
 
 import tensorflow as tf
-from kerastuner.engine import hyperparameters
+from keras_tuner.engine import hyperparameters
 from tensorflow.keras import applications
 from tensorflow.keras import layers
 from tensorflow.python.util import nest
@@ -57,15 +57,15 @@ class DenseBlock(block_module.Block):
     """Block for Dense layers.
 
     # Arguments
-        num_layers: Int or kerastuner.engine.hyperparameters.Choice.
+        num_layers: Int or keras_tuner.engine.hyperparameters.Choice.
             The number of Dense layers in the block.
             If left unspecified, it will be tuned automatically.
-        num_units: Int or kerastuner.engine.hyperparameters.Choice.
+        num_units: Int or keras_tuner.engine.hyperparameters.Choice.
             The number of units in each dense layer.
             If left unspecified, it will be tuned automatically.
         use_bn: Boolean. Whether to use BatchNormalization layers.
             If left unspecified, it will be tuned automatically.
-        dropout: Float or kerastuner.engine.hyperparameters.Choice.
+        dropout: Float or keras_tuner.engine.hyperparameters.Choice.
             The dropout rate for the layers.
             If left unspecified, it will be tuned automatically.
     """
@@ -147,13 +147,13 @@ class RNNBlock(block_module.Block):
     # Arguments
         return_sequences: Boolean. Whether to return the last output in the
             output sequence, or the full sequence. Defaults to False.
-        bidirectional: Boolean or kerastuner.engine.hyperparameters.Boolean.
+        bidirectional: Boolean or keras_tuner.engine.hyperparameters.Boolean.
             Bidirectional RNN. If left unspecified, it will be
             tuned automatically.
-        num_layers: Int or kerastuner.engine.hyperparameters.Choice.
+        num_layers: Int or keras_tuner.engine.hyperparameters.Choice.
             The number of layers in RNN. If left unspecified, it will
             be tuned automatically.
-        layer_type: String or or kerastuner.engine.hyperparameters.Choice.
+        layer_type: String or or keras_tuner.engine.hyperparameters.Choice.
             'gru' or 'lstm'. If left unspecified, it will be tuned
             automatically.
     """
@@ -244,17 +244,17 @@ class ConvBlock(block_module.Block):
     """Block for vanilla ConvNets.
 
     # Arguments
-        kernel_size: Int or kerastuner.engine.hyperparameters.Choice.
+        kernel_size: Int or keras_tuner.engine.hyperparameters.Choice.
             The size of the kernel.
             If left unspecified, it will be tuned automatically.
-        num_blocks: Int or kerastuner.engine.hyperparameters.Choice.
+        num_blocks: Int or keras_tuner.engine.hyperparameters.Choice.
             The number of conv blocks, each of which may contain
             convolutional, max pooling, dropout, and activation. If left unspecified,
             it will be tuned automatically.
         num_layers: Int or hyperparameters.Choice.
             The number of convolutional layers in each block. If left
             unspecified, it will be tuned automatically.
-        filters: Int or kerastuner.engine.hyperparameters.Choice. The number of
+        filters: Int or keras_tuner.engine.hyperparameters.Choice. The number of
             filters in the convolutional layers. If left unspecified, it will
             be tuned automatically.
         max_pooling: Boolean. Whether to use max pooling layer in each block. If left
@@ -487,20 +487,20 @@ class Transformer(block_module.Block):
     # Arguments
         max_features: Int. Size of the vocabulary. Must be set if not using
             TextToIntSequence before this block. Defaults to 20001.
-        pretraining: String or kerastuner.engine.hyperparameters.Choice.
+        pretraining: String or keras_tuner.engine.hyperparameters.Choice.
             'random' (use random weights instead any pretrained
             model), 'glove', 'fasttext' or 'word2vec'. Use pretrained word embedding.
             If left unspecified, it will be tuned automatically.
-        embedding_dim: Int or kerastuner.engine.hyperparameters.Choice.
+        embedding_dim: Int or keras_tuner.engine.hyperparameters.Choice.
             Output dimension of the Attention block.
             If left unspecified, it will be tuned automatically.
-        num_heads: Int or kerastuner.engine.hyperparameters.Choice.
+        num_heads: Int or keras_tuner.engine.hyperparameters.Choice.
             The number of attention heads. If left unspecified,
             it will be tuned automatically.
-        dense_dim: Int or kerastuner.engine.hyperparameters.Choice.
+        dense_dim: Int or keras_tuner.engine.hyperparameters.Choice.
             The output dimension of the Feed-Forward Network. If left
             unspecified, it will be tuned automatically.
-        dropout: Float or kerastuner.engine.hyperparameters.Choice.
+        dropout: Float or keras_tuner.engine.hyperparameters.Choice.
             Between 0 and 1. If left unspecified, it will be
             tuned automatically.
     """
@@ -818,14 +818,14 @@ class Embedding(block_module.Block):
     # Arguments
         max_features: Int. Size of the vocabulary. Must be set if not using
             TextToIntSequence before this block. Defaults to 20001.
-        pretraining: String or kerastuner.engine.hyperparameters.Choice.
+        pretraining: String or keras_tuner.engine.hyperparameters.Choice.
             'random' (use random weights instead any pretrained
             model), 'glove', 'fasttext' or 'word2vec'. Use pretrained word embedding.
             If left unspecified, it will be tuned automatically.
-        embedding_dim: Int or kerastuner.engine.hyperparameters.Choice.
+        embedding_dim: Int or keras_tuner.engine.hyperparameters.Choice.
             Output dimension of the Attention block.
             If left unspecified, it will be tuned automatically.
-        dropout: Float or kerastuner.engine.hyperparameters.Choice.
+        dropout: Float or keras_tuner.engine.hyperparameters.Choice.
             The dropout rate for the layers.
             If left unspecified, it will be tuned automatically.
     """
@@ -929,7 +929,7 @@ class BertBlock(block_module.Block):
         clf = ak.AutoModel(inputs=input_node, outputs=output_node, max_trials=10)
     ```
     # Arguments
-        max_sequence_length: Int or kerastuner.engine.hyperparameters.Choice.
+        max_sequence_length: Int or keras_tuner.engine.hyperparameters.Choice.
             The maximum length of a sequence that is used to train the model.
     """
 

autokeras/engine/named_hypermodel.py

@@ -12,14 +12,14 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import kerastuner
+import keras_tuner
 import tensorflow as tf
 
 from autokeras.engine import serializable
 from autokeras.utils import utils
 
 
-class NamedHyperModel(kerastuner.HyperModel, serializable.Serializable):
+class NamedHyperModel(keras_tuner.HyperModel, serializable.Serializable):
     """
 
     # Arguments

autokeras/engine/tuner.py

@@ -15,9 +15,9 @@
 import copy
 import os
 
-import kerastuner
+import keras_tuner
 import tensorflow as tf
-from kerastuner.engine import hypermodel as hm_module
+from keras_tuner.engine import hypermodel as hm_module
 from tensorflow.keras import callbacks as tf_callbacks
 from tensorflow.keras.layers.experimental import preprocessing
 from tensorflow.python.util import nest
@@ -27,7 +27,7 @@
 from autokeras.utils import utils
 
 
-class AutoTuner(kerastuner.engine.tuner.Tuner):
+class AutoTuner(keras_tuner.engine.tuner.Tuner):
     """A Tuner class based on KerasTuner for AutoKeras.
 
     Different from KerasTuner's Tuner class. AutoTuner's not only tunes the
@@ -41,8 +41,8 @@ class AutoTuner(kerastuner.engine.tuner.Tuner):
     The fully trained model is the best model to be used by AutoModel.
 
     # Arguments
-        oracle: kerastuner Oracle.
-        hypermodel: kerastuner KerasHyperModel.
+        oracle: keras_tuner Oracle.
+        hypermodel: keras_tuner KerasHyperModel.
         **kwargs: The args supported by KerasTuner.
     """
 

autokeras/graph.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import kerastuner
+import keras_tuner
 import tensorflow as tf
 from tensorflow.python.util import nest
 
@@ -51,7 +51,7 @@ def load_graph(filepath, custom_objects=None):
         return Graph.from_config(io_utils.load_json(filepath))
 
 
-class Graph(kerastuner.HyperModel, serializable.Serializable):
+class Graph(keras_tuner.HyperModel, serializable.Serializable):
     """A graph consists of connected Blocks, or Heads.
 
     # Arguments

autokeras/tuners/bayesian_optimization.py

@@ -12,12 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import kerastuner
+import keras_tuner
 
 from autokeras.engine import tuner as tuner_module
 
 
-class BayesianOptimization(kerastuner.BayesianOptimization, tuner_module.AutoTuner):
+class BayesianOptimization(keras_tuner.BayesianOptimization, tuner_module.AutoTuner):
     """KerasTuner BayesianOptimization with preprocessing layer tuning."""
 
     pass

autokeras/tuners/greedy.py

@@ -16,7 +16,7 @@
 from typing import List
 from typing import Optional
 
-import kerastuner
+import keras_tuner
 import numpy as np
 
 from autokeras.engine import tuner as tuner_module
@@ -75,7 +75,7 @@ def get_hp_names(self, node):
         return ret
 
 
-class GreedyOracle(kerastuner.Oracle):
+class GreedyOracle(keras_tuner.Oracle):
     """An oracle combining random search and greedy algorithm.
 
     It groups the HyperParameters into several categories, namely, HyperGraph,
@@ -118,7 +118,7 @@ def _select_hps(self):
         for hp in best_hps.space:
             # Not picking the fixed hps for generating new values.
             if best_hps.is_active(hp) and not isinstance(
-                hp, kerastuner.engine.hyperparameters.Fixed
+                hp, keras_tuner.engine.hyperparameters.Fixed
             ):
                 trie.insert(hp.name)
         all_nodes = trie.nodes
@@ -139,11 +139,11 @@ def _next_initial_hps(self):
                 self._tried_initial_hps[index] = True
                 return hps
 
-    def _populate_space(self, trial_id):
+    def populate_space(self, trial_id):
         if not all(self._tried_initial_hps):
             values = self._next_initial_hps()
             return {
-                "status": kerastuner.engine.trial.TrialStatus.RUNNING,
+                "status": keras_tuner.engine.trial.TrialStatus.RUNNING,
                 "values": values,
             }
 
@@ -155,12 +155,12 @@ def _populate_space(self, trial_id):
                 continue
             # Values found.
             return {
-                "status": kerastuner.engine.trial.TrialStatus.RUNNING,
+                "status": keras_tuner.engine.trial.TrialStatus.RUNNING,
                 "values": values,
             }
         # All stages reached max collisions.
         return {
-            "status": kerastuner.engine.trial.TrialStatus.STOPPED,
+            "status": keras_tuner.engine.trial.TrialStatus.STOPPED,
             "values": None,
         }
 
@@ -176,7 +176,7 @@ def _generate_hp_values(self, hp_names):
 
         collisions = 0
         while True:
-            hps = kerastuner.HyperParameters()
+            hps = keras_tuner.HyperParameters()
             # Generate a set of random values.
             for hp in self.hyperparameters.space:
                 hps.merge([hp])
@@ -207,12 +207,12 @@ def _generate_hp_values(self, hp_names):
 class Greedy(tuner_module.AutoTuner):
     def __init__(
         self,
-        hypermodel: kerastuner.HyperModel,
+        hypermodel: keras_tuner.HyperModel,
         objective: str = "val_loss",
         max_trials: int = 10,
         initial_hps: Optional[List[Dict[str, Any]]] = None,
         seed: Optional[int] = None,
-        hyperparameters: Optional[kerastuner.HyperParameters] = None,
+        hyperparameters: Optional[keras_tuner.HyperParameters] = None,
         tune_new_entries: bool = True,
         allow_new_entries: bool = True,
         **kwargs

autokeras/tuners/hyperband.py

@@ -12,12 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import kerastuner
+import keras_tuner
 
 from autokeras.engine import tuner as tuner_module
 
 
-class Hyperband(kerastuner.Hyperband, tuner_module.AutoTuner):
+class Hyperband(keras_tuner.Hyperband, tuner_module.AutoTuner):
     """KerasTuner Hyperband with preprocessing layer tuning."""
 
     def __init__(

autokeras/tuners/random_search.py

@@ -12,12 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import kerastuner
+import keras_tuner
 
 from autokeras.engine import tuner as tuner_module
 
 
-class RandomSearch(kerastuner.RandomSearch, tuner_module.AutoTuner):
+class RandomSearch(keras_tuner.RandomSearch, tuner_module.AutoTuner):
     """KerasTuner RandomSearch with preprocessing layer tuning."""
 
     pass

autokeras/utils/utils.py

@@ -14,7 +14,7 @@
 
 import re
 
-import kerastuner
+import keras_tuner
 import tensorflow as tf
 from packaging.version import parse
 from tensorflow.python.util import nest
@@ -48,14 +48,14 @@ def check_tf_version() -> None:
 
 
 def check_kt_version() -> None:
-    if parse(kerastuner.__version__) < parse("1.0.2"):
+    if parse(keras_tuner.__version__) < parse("1.0.3"):
         raise ImportError(
-            "The Keras Tuner package version needs to be at least 1.0.2 \n"
+            "The Keras Tuner package version needs to be at least 1.0.3 \n"
             "for AutoKeras to run. Currently, your Keras Tuner version is \n"
             "{version}. Please upgrade with \n"
             "`$ pip install --upgrade keras-tuner`. \n"
             "You can use `pip freeze` to check afterwards that everything is "
-            "ok.".format(version=kerastuner.__version__)
+            "ok.".format(version=keras_tuner.__version__)
         )
 
 
@@ -124,10 +124,10 @@ def add_to_hp(hp, hps, name=None):
     """Add the HyperParameter (self) to the HyperParameters.
 
     # Arguments
-        hp: kerastuner.HyperParameters.
+        hp: keras_tuner.HyperParameters.
         name: String. If left unspecified, the hp name is used.
     """
-    if not isinstance(hp, kerastuner.engine.hyperparameters.HyperParameter):
+    if not isinstance(hp, keras_tuner.engine.hyperparameters.HyperParameter):
         return hp
     kwargs = hp.get_config()
     if name is None:

tests/unit_tests/auto_model_test.py

@@ -14,7 +14,7 @@
 
 from unittest import mock
 
-import kerastuner
+import keras_tuner
 import numpy as np
 import pytest
 import tensorflow as tf
@@ -38,7 +38,7 @@ def test_auto_model_objective_is_kt_objective(tmp_path):
         ak.ImageInput(), ak.RegressionHead(), directory=tmp_path
     )
 
-    assert isinstance(auto_model.objective, kerastuner.Objective)
+    assert isinstance(auto_model.objective, keras_tuner.Objective)
 
 
 def test_auto_model_max_trial_field_as_specified(tmp_path):