- 如何保存模型
保存模型的整个结构+权重+优化器状态
from keras.models import load_model
model.save('my_model.h5')
- 预测
test_pred1 = load_model1.predict([pro1_vec, pro2_vec])
出现报错：
WARNING:tensorflow:5 out of the last 13 calls to
- 当模型带有CUDNN的GRU层
网上有教程说去掉json文件中的cudnn即可，但是我在这一过程中出现了很多报错，可以调用模型的权重然后预测
from tensorflow.keras.models import load_model
def build_model():
***
***
***
return merge_model
model = build_model()
model.load_weights('my_model.h5')
test_pred1 = model([seq1_tensor, seq2_tensor])
参考：
最推荐看这篇：
https://blog.csdn.net/weixin_39653948/article/details/105849959
https://cloud.tencent.com/developer/article/1747486
https://cloud.tencent.com/developer/article/1738224
https://neyzoter.cn/2018/07/14/Keras-Models-Save-Load-Continue-etel/
https://blog.csdn.net/luteresa/article/details/105639675
https://blog.csdn.net/jclian91/article/details/83038861
https://www.icode9.com/content-4-840702.html
https://zhuanlan.zhihu.com/p/272798504
https://www.jianshu.com/p/1c7a9e1c3bd6
https://www.cnblogs.com/tujia/p/13862365.html
https://blog.csdn.net/aaether/article/details/100805874
https://blog.csdn.net/zhou_438/article/details/103391906
https://cloud.tencent.com/developer/ask/sof/266745
https://blog.csdn.net/qq_38978225/article/details/102962084
Original: https://blog.csdn.net/zjy1125/article/details/123598729
Author: zjy1125
Title: CPU调用GPU训练的模型
原创文章受到原创版权保护。转载请注明出处：https://www.johngo689.com/508925/
转载文章受原作者版权保护。转载请注明原作者出处！