Keras用两组数据训练同一个神经网络模型

Keras用两组不同的数据同时训练一个神经网络模型，并将两组数据各自的损失函数显示在进度条后面。该模型的架构如下图所示：

下面是相关的python代码：

import keras
import numpy as np
import tensorflow as tf


loss_tracker0 = keras.metrics.Mean(name="loss0")
loss_tracker1 = keras.metrics.Mean(name="loss1")

mae_metric0   = keras.metrics.MeanAbsoluteError(name="mae0")
mae_metric1   = keras.metrics.MeanAbsoluteError(name="mae1")

class CustomModel(keras.Model):
    def train_step(self, data):
        # Unpack the data. Its structure depends on your model and
        # on what you pass to `fit()`.
        x,y = data

        with tf.GradientTape() as tape:
            y_pred1 = self(x[1], training=True)  # Forward pass
            y_pred0 = self(x[0], training=True)  # Forward pass
            # Compute the loss value
            # (the loss function is configured in `compile()`)
            loss1 = self.compiled_loss(y[1], y_pred1, regularization_losses=self.losses)
            loss0 = self.compiled_loss(y[0], y_pred0, regularization_losses=self.losses)
            
            loss  = loss1+loss0
        # Compute gradients
        trainable_vars = self.trainable_variables
        gradients = tape.gradient(loss, trainable_vars)
        # Update weights
        self.optimizer.apply_gradients(zip(gradients, trainable_vars))

        # Compute our own metrics
        loss_tracker0.update_state(loss0)
        mae_metric0.update_state(y[0], y_pred0)
        
        loss_tracker1.update_state(loss1)
        mae_metric1.update_state(y[1], y_pred1)
        # Return a dict mapping metric names to current value
        return {"loss1": loss_tracker1.result(), "mae1": mae_metric1.result(),"loss0": loss_tracker0.result(), "mae0": mae_metric0.result()}
    
    def test_step(self, data):
        # Unpack the data
        x, y = data
        # Compute predictions
        y_pred1 = self(x[1], training=True)  # Forward pass
        y_pred0 = self(x[0], training=True)  # Forward pass
        # Updates the metrics tracking the loss
        loss1 = self.compiled_loss(y[1], y_pred1, regularization_losses=self.losses)
        loss0 = self.compiled_loss(y[0], y_pred0, regularization_losses=self.losses)
        
        # Compute our own metrics
        loss_tracker0.update_state(loss0)
        mae_metric0.update_state(y[0], y_pred0)
        
        loss_tracker1.update_state(loss1)
        mae_metric1.update_state(y[1], y_pred1)
        
        # Return a dict mapping metric names to current value.
        return {"loss1": loss_tracker1.result(), "mae1": mae_metric1.result(),"loss0": loss_tracker0.result(), "mae0": mae_metric0.result()}

    
# Just use `fit` as usual
x1 = np.random.random((1000, 32))
y1 = np.random.random((1000, 1))

x0 = np.random.random((1000, 32))
y0 = np.random.random((1000, 1))

val_x1 = np.random.random((100, 32))
val_y1 = np.random.random((100, 1))

val_x0 = np.random.random((100, 32))
val_y0 = np.random.random((100, 1))

val_data = ([val_x1,val_x0], [val_y1,val_y0])

# Construct and compile an instance of CustomModel
inputs = keras.Input(shape=(32,))
hidden = keras.layers.Dense(10)(inputs)
outputs = keras.layers.Dense(1)(hidden)
model = CustomModel(inputs, outputs)

model.compile(optimizer="adam", loss=["mse"], metrics=['mae'])

history = model.fit([x1,x0], [y1,y0],validation_data=val_data,verbose=1, epochs=4,batch_size=10)

代码运行的结果如下：