深度学习2.0-21.Keras高层API-compile&fit&Evaluate&Predict

文章目录

• • • 1.标准的训练
    • 4.test
    • 5.predict

1.标准的训练

1.loss and optimizer and metrics

before

now

2.epoch and step

before

now

3.evaluation

before

now

network.compile(optimizer = optimizers.Adam(lr=0.01),
		loss = tf.losses.CategoricalCrossentropy(from_logits = True),
		metrics = ['accuracy']
		)
network.fit(db,epochs=10, validation_data = ds_val,
				validstion_freq=2)

4.test

evaluate相当于中间的test或者validation

5.predict