AutoKeras 3.0 updates (#1951)

* use data.Dataset for all

* some tests passed

* more tests fixed

* more tests fixed.

* image classifier fixed.

* remove data

* text fixed

* removed keras_nlp

* fix one hot encoder.

* More tests fixed

* All tests fixed.

* Add structured data (#1950)

* patch in progress

* patch in progress 2

* progress 3

* tests runnable

* remove tf usages

* structured data working

* all tests passed

* docs works

* structured data added to docs

* clean ups

* Run tests with torch

* add grpc as a dependency

* add grpc and protobuf to test install

* fix the pip install grpcio

* addressing comments.

Author: haifeng-jin

.github/workflows/actions.yml

@@ -10,6 +10,8 @@ jobs:
   build:
     name: Run tests
     runs-on: ubuntu-latest
+    env:
+      KERAS_BACKEND: torch
     steps:
     - uses: actions/checkout@v3
     - name: Set up Python 3.10
@@ -28,7 +30,8 @@ jobs:
         key: ${{ runner.os }}-pip-${{ hashFiles('pyproject.toml') }}
     - name: Install dependencies
       run: |
-        pip install tensorflow
+          pip install torch --index-url https://download.pytorch.org/whl/cpu
+          pip install grpcio protobuf 
         pip install -e ".[tests]" --progress-bar off
     - name: Test with pytest
       run: |
@@ -38,13 +41,17 @@ jobs:
   format:
     name: Check the code format
     runs-on: ubuntu-latest
+    env:
+      KERAS_BACKEND: torch
     steps:
       - uses: actions/checkout@v3
       - name: Run pre-commit
         run: bash shell/pre-commit.sh
   build-docs:
     name: Build the docs
     runs-on: ubuntu-latest
+    env:
+      KERAS_BACKEND: torch
     steps:
       - uses: actions/checkout@v3
       - name: Set up Python 3.10
@@ -54,7 +61,8 @@ jobs:
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip setuptools
-          pip install tensorflow
+          pip install torch --index-url https://download.pytorch.org/whl/cpu
+          pip install grpcio protobuf 
           pip install -e .
           pip install -r docs/requirements.txt
       - name: Build the docs
@@ -66,6 +74,8 @@ jobs:
     needs: [build, format, build-docs]
     if: github.event_name == 'release' && github.event.action == 'created'
     runs-on: ubuntu-latest
+    env:
+      KERAS_BACKEND: torch
     steps:
     - uses: actions/checkout@v3
     - name: Set up Python

autokeras/__init__.py

@@ -12,14 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import keras_nlp
-
 from autokeras.auto_model import AutoModel
-from autokeras.blocks import BertBlock
 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
@@ -29,6 +27,7 @@
 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 XceptionBlock
@@ -39,23 +38,22 @@
 from autokeras.keras_layers import ExpandLastDim
 from autokeras.nodes import ImageInput
 from autokeras.nodes import Input
+from autokeras.nodes import StructuredDataInput
 from autokeras.nodes import TextInput
 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.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
 
-__version__ = "2.1.0dev"
+__version__ = "3.0.0dev"
 
 CUSTOM_OBJECTS = {
-    "BertPreprocessor": keras_nlp.models.BertPreprocessor,
-    "BertBackbone": keras_nlp.models.BertBackbone,
     "CastToFloat32": CastToFloat32,
     "ExpandLastDim": ExpandLastDim,
 }

autokeras/adapters/__init__.py

@@ -14,6 +14,7 @@
 
 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.output_adapters import ClassificationAdapter
 from autokeras.adapters.output_adapters import RegressionAdapter

autokeras/adapters/input_adapters.py

@@ -13,18 +13,17 @@
 # limitations under the License.
 
 import numpy as np
-import tensorflow as tf
 
 from autokeras.engine import adapter as adapter_module
 
 
 class InputAdapter(adapter_module.Adapter):
     def check(self, x):
         """Record any information needed by transform."""
-        if not isinstance(x, (np.ndarray, tf.data.Dataset)):
+        if not isinstance(x, np.ndarray):
             raise TypeError(
-                "Expect the data to Input to be numpy.ndarray or "
-                "tf.data.Dataset, but got {type}.".format(type=type(x))
+                "Expect the data to Input to be numpy.ndarray, "
+                "but got {type}.".format(type=type(x))
             )
         if isinstance(x, np.ndarray) and not np.issubdtype(x.dtype, np.number):
             raise TypeError(
@@ -36,10 +35,10 @@ def check(self, x):
 class ImageAdapter(adapter_module.Adapter):
     def check(self, x):
         """Record any information needed by transform."""
-        if not isinstance(x, (np.ndarray, tf.data.Dataset)):
+        if not isinstance(x, np.ndarray):
             raise TypeError(
-                "Expect the data to ImageInput to be numpy.ndarray or "
-                "tf.data.Dataset, but got {type}.".format(type=type(x))
+                "Expect the data to ImageInput to be numpy.ndarray, "
+                "but got {type}.".format(type=type(x))
             )
         if isinstance(x, np.ndarray) and not np.issubdtype(x.dtype, np.number):
             raise TypeError(
@@ -51,8 +50,17 @@ def check(self, x):
 class TextAdapter(adapter_module.Adapter):
     def check(self, x):
         """Record any information needed by transform."""
-        if not isinstance(x, (np.ndarray, tf.data.Dataset)):
+        if not isinstance(x, np.ndarray):
             raise TypeError(
-                "Expect the data to TextInput to be numpy.ndarray or "
-                "tf.data.Dataset, but got {type}.".format(type=type(x))
+                "Expect the data to TextInput to be numpy.ndarray, "
+                "but got {type}.".format(type=type(x))
+            )
+
+
+class StructuredDataAdapter(adapter_module.Adapter):
+    def check(self, x):
+        if not isinstance(x, np.ndarray):
+            raise TypeError(
+                "Unsupported type {type} for "
+                "{name}.".format(type=type(x), name=self.__class__.__name__)
             )

autokeras/adapters/input_adapters_test.py

@@ -14,8 +14,8 @@
 
 
 import numpy as np
+import pandas as pd
 import pytest
-import tensorflow as tf
 
 from autokeras import test_utils
 from autokeras.adapters import input_adapters
@@ -25,73 +25,70 @@
 def test_image_input_adapter_transform_to_dataset():
     x = test_utils.generate_data()
     adapter = input_adapters.ImageAdapter()
-    assert isinstance(adapter.adapt(x, batch_size=32), tf.data.Dataset)
+    assert isinstance(adapter.adapt(x), np.ndarray)
 
 
 def test_image_input_unsupported_type():
     x = "unknown"
     adapter = input_adapters.ImageAdapter()
     with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
+        x = adapter.adapt(x)
     assert "Expect the data to ImageInput to be numpy" in str(info.value)
 
 
 def test_image_input_numerical():
     x = np.array([[["unknown"]]])
     adapter = input_adapters.ImageAdapter()
     with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
+        x = adapter.adapt(x)
     assert "Expect the data to ImageInput to be numerical" in str(info.value)
 
 
 def test_input_type_error():
     x = "unknown"
     adapter = input_adapters.InputAdapter()
     with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
+        x = adapter.adapt(x)
     assert "Expect the data to Input to be numpy" in str(info.value)
 
 
 def test_input_numerical():
     x = np.array([[["unknown"]]])
     adapter = input_adapters.InputAdapter()
     with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
+        x = adapter.adapt(x)
     assert "Expect the data to Input to be numerical" in str(info.value)
 
 
-def test_text_adapt_unbatched_dataset():
-    x = tf.data.Dataset.from_tensor_slices(np.array(["a b c", "b b c"]))
-    adapter = input_adapters.TextAdapter()
-    x = adapter.adapt(x, batch_size=32)
-
-    assert data_utils.dataset_shape(x).as_list() == [None]
-    assert isinstance(x, tf.data.Dataset)
-
-
-def test_text_adapt_batched_dataset():
-    x = tf.data.Dataset.from_tensor_slices(np.array(["a b c", "b b c"])).batch(
-        32
-    )
-    adapter = input_adapters.TextAdapter()
-    x = adapter.adapt(x, batch_size=32)
-
-    assert data_utils.dataset_shape(x).as_list() == [None]
-    assert isinstance(x, tf.data.Dataset)
-
-
 def test_text_adapt_np():
     x = np.array(["a b c", "b b c"])
     adapter = input_adapters.TextAdapter()
-    x = adapter.adapt(x, batch_size=32)
+    x = adapter.adapt(x)
 
-    assert data_utils.dataset_shape(x).as_list() == [None]
-    assert isinstance(x, tf.data.Dataset)
+    assert data_utils.dataset_shape(x) == [2]
+    assert isinstance(x, np.ndarray)
 
 
 def test_text_input_type_error():
     x = "unknown"
     adapter = input_adapters.TextAdapter()
     with pytest.raises(TypeError) as info:
-        x = adapter.adapt(x, batch_size=32)
+        x = adapter.adapt(x)
     assert "Expect the data to TextInput to be numpy" in str(info.value)
+
+
+def test_structured_data_input_unsupported_type_error():
+    with pytest.raises(TypeError) as info:
+        adapter = input_adapters.StructuredDataAdapter()
+        adapter.adapt("unknown")
+
+    assert "Unsupported type" in str(info.value)
+
+
+def test_structured_data_input_transform_to_dataset():
+    x = pd.read_csv(test_utils.TRAIN_CSV_PATH).to_numpy().astype(str)
+    adapter = input_adapters.StructuredDataAdapter()
+
+    x = adapter.adapt(x)
+
+    assert isinstance(x, np.ndarray)

autokeras/adapters/output_adapters.py

@@ -13,7 +13,6 @@
 # limitations under the License.
 
 import numpy as np
-import tensorflow as tf
 
 from autokeras.engine import adapter as adapter_module
 
@@ -24,10 +23,9 @@ def __init__(self, name, **kwargs):
         self.name = name
 
     def check(self, dataset):
-        supported_types = (tf.data.Dataset, np.ndarray)
-        if not isinstance(dataset, supported_types):
+        if not isinstance(dataset, np.ndarray):
             raise TypeError(
-                f"Expect the target data of {self.name} to be tf.data.Dataset,"
+                f"Expect the target data of {self.name} to be"
                 f" np.ndarray, but got {type(dataset)}."
             )
 

autokeras/adapters/output_adapters_test.py

@@ -14,7 +14,6 @@
 
 import numpy as np
 import pytest
-import tensorflow as tf
 
 from autokeras.adapters import output_adapters
 
@@ -23,14 +22,14 @@ def test_unsupported_types_error():
     adapter = output_adapters.ClassificationAdapter(name="a")
 
     with pytest.raises(TypeError) as info:
-        adapter.adapt(1, batch_size=32)
+        adapter.adapt(1)
 
-    assert "Expect the target data of a to be tf" in str(info.value)
+    assert "Expect the target data of a to be" in str(info.value)
 
 
 def test_reg_head_transform_1d_np():
     adapter = output_adapters.RegressionAdapter(name="a")
 
-    y = adapter.adapt(np.random.rand(10), batch_size=32)
+    y = adapter.adapt(np.random.rand(10))
 
-    assert isinstance(y, tf.data.Dataset)
+    assert isinstance(y, np.ndarray)

autokeras/analysers/__init__.py

@@ -12,8 +12,11 @@
 # 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.output_analysers import ClassificationAnalyser
 from autokeras.analysers.output_analysers import RegressionAnalyser