import tensorflow as tf
import numpy as np
from sklearn.datasets import load_iris

data = load_iris()
iris_data = np.float32(data.data)
iris_target = np.float32(data.target)
iris_target = np.float32(tf.keras.utils.to_categorical(iris_target, num_classes=3))
train_data = tf.data.Dataset.from_tensor_slices((iris_data, iris_target)).batch(128)


class MyLayer(tf.keras.layers.Layer):
    def __init__(self, output_dim):
        self.output_dim = output_dim
        super(MyLayer, self).__init__()

    def build(self, input_shape):
        self.weight = tf.Variable(tf.random.normal([input_shape[-1], self.output_dim]), name="dense_weight")
        self.bias = tf.Variable(tf.random.normal([self.output_dim]), name="bias_weight")
        super(MyLayer, self).build(input_shape)  # Be sure to call this somewhere!

    def call(self, input_tensor):
        out = tf.matmul(input_tensor, self.weight) + self.bias
        out = tf.nn.relu(out)
        out = tf.keras.layers.Dropout(0.1)(out)
        return out


input_xs = tf.keras.Input(shape=(4), name='input_xs')
out = tf.keras.layers.Dense(32, activation='relu', name='dense_1')(input_xs)
out = MyLayer(32)(out)
out = MyLayer(48)(out)
out = tf.keras.layers.Dense(64, activation='relu', name='dense_2')(out)
logits = tf.keras.layers.Dense(3, activation="softmax", name='predictions')(out)
model = tf.keras.Model(inputs=input_xs, outputs=logits)
print(model.summary())