tensorFlow 2.0 如何加载checkpoint

最新推荐文章于 2022-10-08 11:00:19 发布

Firesuiry

最新推荐文章于 2022-10-08 11:00:19 发布

阅读量3.2k

点赞数 3

分类专栏： autoLOL tensorflow2

本文链接：https://blog.youkuaiyun.com/weixin_44387667/article/details/103001355

版权

解决了一个令人头秃的问题


    early_stopping=tf.keras.callbacks.EarlyStopping(monitor='loss', min_delta=0.0001,
                                  patience=5, mode='auto',
                                  baseline=None, restore_best_weights=False,verbose=1)
    reduce_lr = tf.keras.callbacks.ReduceLROnPlateau(monitor='loss',patience=1,verbose=1,factor = 0.3)
    checkpoint = tf.keras.callbacks.ModelCheckpoint('model_cp/',monitor='loss', 
                                                    save_weights_only=False,verbose=1,
                                                    save_best_only=True,save_freq='epoch',mode = 'min')
    new_model.fit(ds, epochs=10000, steps_per_epoch=steps_per_epoch,
                  callbacks = [early_stopping,reduce_lr,checkpoint])

训练代码长这样子

自动保