Tensorflow2.0 confusion matrix does not match printing accuracy

Recently working on a binary classification project, the last few layers of the model are like this, and finally activated with sigmiod.

model.add(tf.keras.layers.Dense(64, activation='relu'))model.add(tf.keras.layers.Dropout(0.5)),model.add(tf.keras.layers.Dense(1, activation='sigmoid'))

The loss function uses this

tf.keras.losses.binary_crossentropy

Just run it

model.fit(...)

The effect of the printed information is not bad. Finally, I took a look at the confusion matrix again, and I vomited.In this way, it is OK to use y_pred.

pred = self.model.predict(x)y_pred = (pred>0.5).astype(int)