Support Keras 3 with TF backend (#1908)

* initial renaming

* renaming tree

* rename tensorflow.keras

* update

* update version numbers

* update imports to keras

* now the tests can run bug failed some

* fix more tests

* fix more tests

* fix more tests

* removed structured data and time series, object detection, and segmentation

* remove structured data from docs

* fix docs

* minor fix

* fix more tests

* fix more tests

* fix more tests

* update docs

* update

* fix more tests

* add notes & improve coverage

* fix tests

* update

* update

* update input block

* fix coverage

* update the tutorials

RELEASE.md

@@ -1,3 +1,25 @@
+# Release v2.0.0
+
+## Breaking changes
+
+* Requires `keras>=3.0.0` instead of `tf.keras`.
+* Removed the structured data related tasks by removing the following public
+  APIs:
+  * `CategoricalToNumerical`
+  * `MultiCategoryEncoding`
+  * `StructuredDataInput`
+  * `StructuredDataBlock`
+  * `StructuredDataClassifier`
+  * `StructuredDataRegressor`
+* Removed the Time series related tasks by removing the following public APIs:
+  * `TimeseriesInput`
+  * `TimeseriesForecaster`
+* Reduced search space of Text related tasks by removing the following blocks.
+  * `Embedding`
+  * `TextToIntSequence`
+  * `TextToNgramVector`
+  * `Transformer`
+
 # Release v1.1.0
 
 ## Breaking changes

autokeras/__init__.py

@@ -16,12 +16,10 @@
 
 from autokeras.auto_model import AutoModel
 from autokeras.blocks import BertBlock
-from autokeras.blocks import CategoricalToNumerical
 from autokeras.blocks import ClassificationHead
 from autokeras.blocks import ConvBlock
 from autokeras.blocks import DenseBlock
 from autokeras.blocks import EfficientNetBlock
-from autokeras.blocks import Embedding
 from autokeras.blocks import Flatten
 from autokeras.blocks import ImageAugmentation
 from autokeras.blocks import ImageBlock
@@ -31,49 +29,33 @@
 from autokeras.blocks import ResNetBlock
 from autokeras.blocks import RNNBlock
 from autokeras.blocks import SpatialReduction
-from autokeras.blocks import StructuredDataBlock
 from autokeras.blocks import TemporalReduction
 from autokeras.blocks import TextBlock
-from autokeras.blocks import TextToIntSequence
-from autokeras.blocks import TextToNgramVector
-from autokeras.blocks import Transformer
 from autokeras.blocks import XceptionBlock
 from autokeras.engine.block import Block
 from autokeras.engine.head import Head
 from autokeras.engine.node import Node
 from autokeras.keras_layers import CastToFloat32
 from autokeras.keras_layers import ExpandLastDim
-from autokeras.keras_layers import MultiCategoryEncoding
 from autokeras.nodes import ImageInput
 from autokeras.nodes import Input
-from autokeras.nodes import StructuredDataInput
 from autokeras.nodes import TextInput
-from autokeras.nodes import TimeseriesInput
 from autokeras.tasks import ImageClassifier
 from autokeras.tasks import ImageRegressor
-from autokeras.tasks import StructuredDataClassifier
-from autokeras.tasks import StructuredDataRegressor
 from autokeras.tasks import TextClassifier
 from autokeras.tasks import TextRegressor
-from autokeras.tasks import TimeseriesForecaster
 from autokeras.tuners import BayesianOptimization
 from autokeras.tuners import Greedy
 from autokeras.tuners import Hyperband
 from autokeras.tuners import RandomSearch
 from autokeras.utils.io_utils import image_dataset_from_directory
 from autokeras.utils.io_utils import text_dataset_from_directory
-from autokeras.utils.utils import check_kt_version
-from autokeras.utils.utils import check_tf_version
 
-__version__ = "1.1.1dev"
-
-check_tf_version()
-check_kt_version()
+__version__ = "2.0.0dev"
 
 CUSTOM_OBJECTS = {
     "BertPreprocessor": keras_nlp.models.BertPreprocessor,
     "BertBackbone": keras_nlp.models.BertBackbone,
     "CastToFloat32": CastToFloat32,
     "ExpandLastDim": ExpandLastDim,
-    "MultiCategoryEncoding": MultiCategoryEncoding,
 }

autokeras/adapters/__init__.py

@@ -14,9 +14,6 @@
 
 from autokeras.adapters.input_adapters import ImageAdapter
 from autokeras.adapters.input_adapters import InputAdapter
-from autokeras.adapters.input_adapters import StructuredDataAdapter
 from autokeras.adapters.input_adapters import TextAdapter
-from autokeras.adapters.input_adapters import TimeseriesAdapter
 from autokeras.adapters.output_adapters import ClassificationAdapter
 from autokeras.adapters.output_adapters import RegressionAdapter
-from autokeras.adapters.output_adapters import SegmentationHeadAdapter

autokeras/adapters/input_adapters.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import pandas as pd
 import tensorflow as tf
 
 from autokeras.engine import adapter as adapter_module
@@ -57,40 +56,3 @@ def check(self, x):
                 "Expect the data to TextInput to be numpy.ndarray or "
                 "tf.data.Dataset, but got {type}.".format(type=type(x))
             )
-
-
-class StructuredDataAdapter(adapter_module.Adapter):
-    def check(self, x):
-        if not isinstance(x, (pd.DataFrame, np.ndarray, tf.data.Dataset)):
-            raise TypeError(
-                "Unsupported type {type} for "
-                "{name}.".format(type=type(x), name=self.__class__.__name__)
-            )
-
-    def convert_to_dataset(self, dataset, batch_size):
-        if isinstance(dataset, pd.DataFrame):
-            dataset = dataset.values
-        if isinstance(dataset, np.ndarray) and dataset.dtype == object:
-            dataset = dataset.astype(str)
-        return super().convert_to_dataset(dataset, batch_size)
-
-
-class TimeseriesAdapter(adapter_module.Adapter):
-    def __init__(self, lookback=None, **kwargs):
-        super().__init__(**kwargs)
-        self.lookback = lookback
-
-    def check(self, x):
-        """Record any information needed by transform."""
-        if not isinstance(x, (pd.DataFrame, np.ndarray, tf.data.Dataset)):
-            raise TypeError(
-                "Expect the data in TimeseriesInput to be numpy.ndarray"
-                " or tf.data.Dataset or pd.DataFrame, but got {type}.".format(
-                    type=type(x)
-                )
-            )
-
-    def convert_to_dataset(self, dataset, batch_size):
-        if isinstance(dataset, pd.DataFrame):
-            dataset = dataset.values
-        return super().convert_to_dataset(dataset, batch_size)

autokeras/adapters/input_adapters_test.py

@@ -14,7 +14,6 @@
 
 
 import numpy as np
-import pandas as pd
 import pytest
 import tensorflow as tf
 
@@ -23,25 +22,6 @@
 from autokeras.utils import data_utils
 
 
-def test_structured_data_input_unsupported_type_error():
-    with pytest.raises(TypeError) as info:
-        adapter = input_adapters.StructuredDataAdapter()
-        adapter.adapt("unknown", batch_size=32)
-
-    assert "Unsupported type" in str(info.value)
-
-
-def test_structured_data_input_transform_to_dataset():
-    x = tf.data.Dataset.from_tensor_slices(
-        pd.read_csv(test_utils.TRAIN_CSV_PATH).to_numpy().astype(str)
-    )
-    adapter = input_adapters.StructuredDataAdapter()
-
-    x = adapter.adapt(x, batch_size=32)
-
-    assert isinstance(x, tf.data.Dataset)
-
-
 def test_image_input_adapter_transform_to_dataset():
     x = test_utils.generate_data()
     adapter = input_adapters.ImageAdapter()
@@ -115,19 +95,3 @@ def test_text_input_type_error():
     with pytest.raises(TypeError) as info:
         x = adapter.adapt(x, batch_size=32)
     assert "Expect the data to TextInput to be numpy" in str(info.value)
-
-
-def test_time_series_input_type_error():
-    x = "unknown"
-    adapter = input_adapters.TimeseriesAdapter()
-    with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
-    assert "Expect the data in TimeseriesInput to be numpy" in str(info.value)
-
-
-def test_time_series_input_transform_df_to_dataset():
-    adapter = input_adapters.TimeseriesAdapter()
-
-    x = adapter.adapt(pd.DataFrame(np.random.rand(100, 32)), batch_size=32)
-
-    assert isinstance(x, tf.data.Dataset)

autokeras/adapters/output_adapters.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import pandas as pd
 import tensorflow as tf
 
 from autokeras.engine import adapter as adapter_module
@@ -25,19 +24,14 @@ def __init__(self, name, **kwargs):
         self.name = name
 
     def check(self, dataset):
-        supported_types = (tf.data.Dataset, np.ndarray, pd.DataFrame, pd.Series)
+        supported_types = (tf.data.Dataset, np.ndarray)
         if not isinstance(dataset, supported_types):
             raise TypeError(
                 f"Expect the target data of {self.name} to be tf.data.Dataset,"
-                f" np.ndarray, pd.DataFrame or pd.Series, "
-                f"but got {type(dataset)}."
+                f" np.ndarray, but got {type(dataset)}."
             )
 
     def convert_to_dataset(self, dataset, batch_size):
-        if isinstance(dataset, pd.DataFrame):
-            dataset = dataset.values
-        if isinstance(dataset, pd.Series):
-            dataset = dataset.values
         return super().convert_to_dataset(dataset, batch_size)
 
 
@@ -47,7 +41,3 @@ class ClassificationAdapter(HeadAdapter):
 
 class RegressionAdapter(HeadAdapter):
     pass
-
-
-class SegmentationHeadAdapter(ClassificationAdapter):
-    pass

autokeras/adapters/output_adapters_test.py

@@ -13,37 +13,12 @@
 # limitations under the License.
 
 import numpy as np
-import pandas as pd
 import pytest
 import tensorflow as tf
 
-from autokeras import test_utils
 from autokeras.adapters import output_adapters
 
 
-def test_clf_head_transform_pd_series_to_dataset():
-    adapter = output_adapters.ClassificationAdapter(name="a")
-
-    y = adapter.adapt(
-        pd.read_csv(test_utils.TEST_CSV_PATH).pop("survived"), batch_size=32
-    )
-
-    assert isinstance(y, tf.data.Dataset)
-
-
-def test_clf_head_transform_df_to_dataset():
-    adapter = output_adapters.ClassificationAdapter(name="a")
-
-    y = adapter.adapt(
-        pd.DataFrame(
-            test_utils.generate_one_hot_labels(dtype="np", num_classes=10)
-        ),
-        batch_size=32,
-    )
-
-    assert isinstance(y, tf.data.Dataset)
-
-
 def test_unsupported_types_error():
     adapter = output_adapters.ClassificationAdapter(name="a")
 
@@ -53,16 +28,6 @@ def test_unsupported_types_error():
     assert "Expect the target data of a to be tf" in str(info.value)
 
 
-def test_reg_head_transform_pd_series():
-    adapter = output_adapters.RegressionAdapter(name="a")
-
-    y = adapter.adapt(
-        pd.read_csv(test_utils.TEST_CSV_PATH).pop("survived"), batch_size=32
-    )
-
-    assert isinstance(y, tf.data.Dataset)
-
-
 def test_reg_head_transform_1d_np():
     adapter = output_adapters.RegressionAdapter(name="a")
 

autokeras/analysers/__init__.py

@@ -12,12 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-from autokeras.analysers.input_analysers import CATEGORICAL
-from autokeras.analysers.input_analysers import NUMERICAL
 from autokeras.analysers.input_analysers import ImageAnalyser
 from autokeras.analysers.input_analysers import InputAnalyser
-from autokeras.analysers.input_analysers import StructuredDataAnalyser
 from autokeras.analysers.input_analysers import TextAnalyser
-from autokeras.analysers.input_analysers import TimeseriesAnalyser
 from autokeras.analysers.output_analysers import ClassificationAnalyser
 from autokeras.analysers.output_analysers import RegressionAnalyser