tensoflow2

版本变化

高效的TensorFlow 2.0（tensorflow2官方教程翻译）
TensorFlow2.0 问世，Pytorch还能否撼动老大哥地位

tensorflow各版本离线下载网站

1 环境要求

CUDA10

2 安装命令

参考：https://blog.csdn.net/u011119817/article/details/88309256

pip install tensorflow==2.0.0b1                  #cpu版本
pip install tensorflow-gpu==2.0.0b1          #gpu版本

pip install tensorflow==2.0.0rc2                  #cpu版本
pip install tensorflow-gpu==2.0.0rc2          #gpu版本

3 debug

#x_train is tensor
tf.print('tf.shape(x_train):',tf.shape(x_train))
tf.print('tf.shape(y_train):',tf.shape(y_train))
tf.print(x_train)

4 mult gpu

参考：https://www.jianshu.com/p/d57595dac5a9

5.两步轻松实现在Keras中使用Tensorboard

参考：https://www.lizenghai.com/archives/5135.html

tensorboard --logdir logs

6.keras保存和恢复模型

https://www.cnblogs.com/augustone/p/10507185.html

7.BatchNormalization坑

http://www.cainiaoxueyuan.com/suanfa/11644.html

8.问题:default/dso_loader.cc:53] Could not dlopen library ‘cupti64_100.dll’; dlerror: cupti64_100.dll not found

参考：https://github.com/tensorflow/tensorflow/issues/6235
解决方法：
I have encountered this problem before. When you use CUDA 8.0,the file cupti64_80.dll lies in C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\extras\CUPTI\libx64. I just fixed the problem by copying the dll into C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin, and the file cupti.lib in the same location into C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64. And it works!

1.复制
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\extras\CUPTI\libx64\cupti64_80.dll
到
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\bin\

2.复制
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\extras\CUPTI\libx64\cupti.lib
到
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0\lib\x64\