如何在 Keras 训练期间保存动力？
[英]How to save momentum during training in Keras?
问题描述
I want to be able to save the state of training my neural net using SGD.
我希望能够使用 SGD 保存训练我的神经网络的状态。 I know I can save the weights using the save_weights function.我知道我可以使用 save_weights 函数保存权重。 What I don't know is how to save the momentum of the gradient decent.我不知道的是如何保存梯度体面的动量。 Does anyone know how to do this?有谁知道如何做到这一点？
1 个解决方案
解决方案1
1 2016-05-11 04:59:11
使用model.optimizer.get_state()和model.optimizer.set_state() - 请参阅https://github.com/fchollet/keras/issues/454 。
声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.
相关问题
如何在Keras中保存训练期间保存的模型列表（仅保存最佳）？ - How to keep a list of models saved during training (save best only) in Keras?
如何在Keras训练期间记录张量的值？ - How to record the values of a tensor during training in Keras?
如何在 Keras 的每个时期保存训练历史？ - How to save training history on every epoch in Keras?
Keras：如何保存 model 并继续训练？ - Keras: How to save model and continue training?
如何在训练期间替换损失函数 tensorflow.keras - How to replace loss function during training tensorflow.keras
如何在训练自动编码器（回调）期间将 keras 中的输入随机设置为零？ - How to randomly set inputs to zero in keras during training autoencoder (callback)?
如何在keras训练期间调用reset_states（）？ - How to call reset_states() during training in keras?
在 Keras 训练期间如何在损失函数内打印？ - How do I print inside the loss function during training in Keras?
如何防止 Keras 在训练期间计算指标 - How to prevent Keras from computing metrics during training
如何在 tf.keras 训练期间获得当前时代的进度？ - How to get progress of current epoch during tf.keras training?
相关标签