parameters(), args.lr, momentum=args.momentum, weight_decay=args.weight_decay) scheduler = ReduceLROnPlateau(optimizer, 'min') for epoch in xrange(args.start_epoch, args.epochs): train(train_loader 4/0b74d452a3986ba46ffa8dea87586ace/p10_1.jpg) ## Scheme 1 CLASS torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, mode='min', factor=0.1, patience=10, verbose=False, threshold=0.0001, th parameters(), args.lr, momentum=args.momentum, weight_decay=args.weight_decay) scheduler = ReduceLROnPlateau(optimizer, 'min') for epoch in xrange(args.start_epoch, args.epochs): train(train_loader

148 11.1.9 LearningRateScheduler [source] 148 11.1.10 TensorBoard [source] 149 11.1.11 ReduceLROnPlateau [source] 149 11.1.12 CSVLogger [source] 150 11.1.13 LambdaCallback [source] 150 11.2 创建一个回调函数 一个字典，对应层的名字到保存有这个嵌入层元数据文件的名字。查看详情关于元数据的数据格式。以防同样的元数据被用于所用的嵌入层，字符串可以被传入。 #### 11.1.11 ReduceLROnPlateau [source] keras.callbacks.ReduceLROnPlateau(monitor='val_loss', factor=0.1, patience=10, verbose=0, mode='auto' 当学习停止时，模型总是会受益于降低 2-10 倍的学习速率。这个回调函数监测一个数据并且当这个数据在一定「有耐心」的训练轮之后还没有进步，那么学习速率就会被降低。 ## 例 reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.2, patience=5, min_lr=0.001) model.fit(X_train, Y_train