Add more blocks to autoencoders

scico/flax/autoencoders/blocks.py

@@ -19,6 +19,7 @@
 import flax.linen as nn
 from flax.core import Scope  # noqa
 from flax.linen.module import _Sentinel  # noqa
+from scico.flax.blocks import upscale_nn
 
 # The imports of Scope and _Sentinel (above) are required to silence
 # "cannot resolve forward reference" warnings when building sphinx api
@@ -121,3 +122,84 @@ def __call__(self, x: ArrayLike) -> ArrayLike:
             x = x.reshape((x.shape[0], -1))
 
         return x
+
+
+class ConvPoolBlock(nn.Module):
+    """Define convolution and pooling Flax block.
+
+    Args:
+        num_filters: Number of filters in the convolutional layer of the
+            block. Corresponds to the number of channels in the output
+            tensor.
+        kernel_size: A shape tuple defining the size of the convolution
+            filters.
+        strides: A shape tuple defining the size of strides in convolution.
+        activation: Flax function defining the activation operation to apply.
+        pooling: Flax function defining the pooling operation to apply.
+        window_shape: A shape tuple defining the window to reduce over in
+            the pooling operation.
+    """
+
+    num_filters: int
+    kernel_size: Tuple[int, int] = (3, 3)
+    strides: Tuple[int, int] = (1, 1)
+    activation: Callable = nn.leaky_relu
+    pooling: Callable = nn.max_pool
+    window_shape: Tuple[int, int] = (2, 2)
+
+    @nn.compact
+    def __call__(self, x: ArrayLike) -> ArrayLike:
+        """Apply convolution followed by activation and pooling.
+
+        Args:
+            inputs: The array to be transformed.
+
+        Returns:
+            The transformed input.
+        """
+        x = nn.Conv(
+            self.num_filters,
+            self.kernel_size,
+            strides=self.strides,
+            use_bias=False,
+            padding="CIRCULAR",
+        )(x)
+        x = self.activation(x)
+        x = self.pooling(x, self.window_shape, strides=self.window_shape, padding="SAME")
+
+        return x
+
+
+class ConvUpsampleBlock(nn.Module):
+    """Define convolution and upsample Flax block.
+
+    Args:
+        num_filters: Number of filters in the convolutional layer of the
+            block. Corresponds to the number of channels in the output
+            tensor.
+        kernel_size: A shape tuple defining the size of the convolution
+            filters.
+        strides: A shape tuple defining the size of strides in convolution.
+        activation: Flax function defining the activation operation to apply.
+        upsampling_scale: Integer scaling factor.
+    """
+
+    num_filters: int
+    kernel_size: Tuple[int, int] = (3, 3)
+    strides: Tuple[int, int] = (1, 1)
+    activation: Callable = nn.leaky_relu
+    upsampling_scale: int = 2
+
+    @nn.compact
+    def __call__(self, x: ArrayLike) -> ArrayLike:
+        x = nn.ConvTranspose(
+            self.num_filters,
+            self.kernel_size,
+            strides=self.strides,
+            use_bias=False,
+            padding="CIRCULAR",
+        )(x)
+        x = self.activation(x)
+        x = upscale_nn(x, self.upsampling_scale)
+
+        return x