[英]How to retrain model in graph (.pb)?
I have model saved in graph (.pb file). 我将模型保存在图形(.pb文件)中。 But now the model is inaccurate and I would like to develop it.
但是现在该模型不准确,我想开发它。 I have pictures of additional data to learn, but I don't if it's possible or if it's how to do it?
我有要学习的其他数据的图片,但是如果有可能或者它是怎么做的,我没有吗? The result must be the modified of new data pb graph.
结果必须是新数据pb图的修改。
It's a good question. 这是一个好问题。 Actually it would be nice, if someone could explain how to do this.
实际上,如果有人可以解释如何做到这一点,那就太好了。 But in addition i can say you, that it would come to "catastrophic forgetting", so it wouldn't work out.
但除此之外,我可以对您说,这将导致“灾难性的遗忘”,因此无法解决。 You had to train all your data again.
您必须再次训练所有数据。
But anyway, i also would like to know that espacially for ssd, just for test reasons. 但是无论如何,我也想出于测试的目的,特别针对ssd知道这一点。
The mozzila/DeepSpeech community has contributed a way to initialize training from a frozen graph(.pb). mozzila / DeepSpeech社区提供了一种从冻结图(.pb)初始化训练的方法。 It does not restores optimizer parameters, so adjusting the learning rate is necessary.
它不会恢复优化器参数,因此需要调整学习率。
You could find the code at: 您可以在以下位置找到代码:
https://github.com/mozilla/DeepSpeech/blob/master/DeepSpeech.py#L1562 https://github.com/mozilla/DeepSpeech/blob/master/DeepSpeech.py#L1562
Hope this helps! 希望这可以帮助!
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.