Add ConvMixer

kimm/models/__init__.py

@@ -1,4 +1,5 @@
 from kimm.models.base_model import BaseModel
+from kimm.models.convmixer import *  # noqa:F403
 from kimm.models.densenet import *  # noqa:F403
 from kimm.models.efficientnet import *  # noqa:F403
 from kimm.models.ghostnet import *  # noqa:F403

kimm/models/convmixer.py

@@ -0,0 +1,285 @@
+import typing
+
+import keras
+from keras import layers
+from keras import utils
+
+from kimm.models.base_model import BaseModel
+from kimm.utils import add_model_to_registry
+
+
+def apply_convmixer_block(
+    inputs, output_channels, kernel_size, activation, name="convmixer_block"
+):
+    x = inputs
+
+    # Depthwise
+    x = layers.DepthwiseConv2D(
+        kernel_size,
+        1,
+        padding="same",
+        activation=activation,
+        use_bias=True,
+        name=f"{name}_0_fn_0_dwconv2d",
+    )(x)
+    x = layers.BatchNormalization(
+        momentum=0.9, epsilon=1e-5, name=f"{name}_0_fn_2"
+    )(x)
+    x = layers.Add()([x, inputs])
+
+    # Pointwise
+    x = layers.Conv2D(
+        output_channels,
+        1,
+        1,
+        activation=activation,
+        use_bias=True,
+        name=f"{name}_1_conv2d",
+    )(x)
+    x = layers.BatchNormalization(momentum=0.9, epsilon=1e-5, name=f"{name}_3")(
+        x
+    )
+    return x
+
+
+class ConvMixer(BaseModel):
+    def __init__(
+        self,
+        depth: int = 32,
+        hidden_channels: int = 768,
+        patch_size: int = 7,
+        kernel_size: int = 7,
+        activation: str = "relu",
+        **kwargs,
+    ):
+        parsed_kwargs = self.parse_kwargs(kwargs)
+        img_input = self.determine_input_tensor(
+            parsed_kwargs["input_tensor"],
+            parsed_kwargs["input_shape"],
+            parsed_kwargs["default_size"],
+        )
+        x = img_input
+
+        if parsed_kwargs["include_preprocessing"]:
+            x = self.build_preprocessing(x, "imagenet")
+
+        # Prepare feature extraction
+        features = {}
+
+        # Stem
+        x = layers.Conv2D(
+            hidden_channels,
+            patch_size,
+            patch_size,
+            activation=activation,
+            use_bias=True,
+            name="stem_conv2d",
+        )(x)
+        x = layers.BatchNormalization(
+            momentum=0.9, epsilon=1e-5, name="stem_bn"
+        )(x)
+        features["STEM"] = x
+
+        # Blocks
+        for i in range(depth):
+            x = apply_convmixer_block(
+                x, hidden_channels, kernel_size, activation, name=f"blocks_{i}"
+            )
+            # Add feature
+            features[f"BLOCK{i}"] = x
+
+        # Head
+        if parsed_kwargs["include_top"]:
+            x = self.build_top(
+                x,
+                parsed_kwargs["classes"],
+                parsed_kwargs["classifier_activation"],
+                parsed_kwargs["dropout_rate"],
+            )
+        else:
+            if parsed_kwargs["pooling"] == "avg":
+                x = layers.GlobalAveragePooling2D(name="avg_pool")(x)
+            elif parsed_kwargs["pooling"] == "max":
+                x = layers.GlobalMaxPooling2D(name="max_pool")(x)
+
+        # Ensure that the model takes into account
+        # any potential predecessors of `input_tensor`.
+        if parsed_kwargs["input_tensor"] is not None:
+            inputs = utils.get_source_inputs(parsed_kwargs["input_tensor"])
+        else:
+            inputs = img_input
+
+        super().__init__(inputs=inputs, outputs=x, features=features, **kwargs)
+
+        # All references to `self` below this line
+        self.add_references(parsed_kwargs)
+        self.depth = depth
+        self.hidden_channels = hidden_channels
+        self.patch_size = patch_size
+        self.kernel_size = kernel_size
+        self.activation = activation
+
+    @staticmethod
+    def available_feature_keys():
+        raise NotImplementedError
+
+    def get_config(self):
+        config = super().get_config()
+        config.update(
+            {
+                "depth": self.depth,
+                "hidden_channels": self.hidden_channels,
+                "patch_size": self.patch_size,
+                "kernel_size": self.kernel_size,
+                "activation": self.activation,
+            }
+        )
+        return config
+
+    def fix_config(self, config):
+        unused_kwargs = [
+            "depth",
+            "hidden_channels",
+            "patch_size",
+            "kernel_size",
+            "activation",
+        ]
+        for k in unused_kwargs:
+            config.pop(k, None)
+        return config
+
+
+"""
+Model Definition
+"""
+
+
+class ConvMixer736D32(ConvMixer):
+    def __init__(
+        self,
+        input_tensor: keras.KerasTensor = None,
+        input_shape: typing.Optional[typing.Sequence[int]] = None,
+        include_preprocessing: bool = True,
+        include_top: bool = True,
+        pooling: typing.Optional[str] = None,
+        dropout_rate: float = 0.0,
+        classes: int = 1000,
+        classifier_activation: str = "softmax",
+        weights: typing.Optional[str] = None,
+        name: str = "ConvMixer736D32",
+        **kwargs,
+    ):
+        kwargs = self.fix_config(kwargs)
+        super().__init__(
+            32,
+            768,
+            7,
+            7,
+            "relu",
+            input_tensor=input_tensor,
+            input_shape=input_shape,
+            include_preprocessing=include_preprocessing,
+            include_top=include_top,
+            pooling=pooling,
+            dropout_rate=dropout_rate,
+            classes=classes,
+            classifier_activation=classifier_activation,
+            weights=weights,
+            name=name,
+            **kwargs,
+        )
+
+    @staticmethod
+    def available_feature_keys():
+        feature_keys = ["STEM"]
+        feature_keys.extend([f"BLOCK{i}" for i in range(32)])
+        return feature_keys
+
+
+class ConvMixer1024D20(ConvMixer):
+    def __init__(
+        self,
+        input_tensor: keras.KerasTensor = None,
+        input_shape: typing.Optional[typing.Sequence[int]] = None,
+        include_preprocessing: bool = True,
+        include_top: bool = True,
+        pooling: typing.Optional[str] = None,
+        dropout_rate: float = 0.0,
+        classes: int = 1000,
+        classifier_activation: str = "softmax",
+        weights: typing.Optional[str] = None,
+        name: str = "ConvMixer1024D20",
+        **kwargs,
+    ):
+        kwargs = self.fix_config(kwargs)
+        super().__init__(
+            20,
+            1024,
+            14,
+            9,
+            "gelu",
+            input_tensor=input_tensor,
+            input_shape=input_shape,
+            include_preprocessing=include_preprocessing,
+            include_top=include_top,
+            pooling=pooling,
+            dropout_rate=dropout_rate,
+            classes=classes,
+            classifier_activation=classifier_activation,
+            weights=weights,
+            name=name,
+            **kwargs,
+        )
+
+    @staticmethod
+    def available_feature_keys():
+        feature_keys = ["STEM"]
+        feature_keys.extend([f"BLOCK{i}" for i in range(20)])
+        return feature_keys
+
+
+class ConvMixer1536D20(ConvMixer):
+    def __init__(
+        self,
+        input_tensor: keras.KerasTensor = None,
+        input_shape: typing.Optional[typing.Sequence[int]] = None,
+        include_preprocessing: bool = True,
+        include_top: bool = True,
+        pooling: typing.Optional[str] = None,
+        dropout_rate: float = 0.0,
+        classes: int = 1000,
+        classifier_activation: str = "softmax",
+        weights: typing.Optional[str] = None,
+        name: str = "ConvMixer1536D20",
+        **kwargs,
+    ):
+        kwargs = self.fix_config(kwargs)
+        super().__init__(
+            20,
+            1536,
+            7,
+            9,
+            "gelu",
+            input_tensor=input_tensor,
+            input_shape=input_shape,
+            include_preprocessing=include_preprocessing,
+            include_top=include_top,
+            pooling=pooling,
+            dropout_rate=dropout_rate,
+            classes=classes,
+            classifier_activation=classifier_activation,
+            weights=weights,
+            name=name,
+            **kwargs,
+        )
+
+    @staticmethod
+    def available_feature_keys():
+        feature_keys = ["STEM"]
+        feature_keys.extend([f"BLOCK{i}" for i in range(20)])
+        return feature_keys
+
+
+add_model_to_registry(ConvMixer736D32, "imagenet")
+add_model_to_registry(ConvMixer1024D20, "imagenet")
+add_model_to_registry(ConvMixer1536D20, "imagenet")