Tensorflow2疑难问题---2、tensorflow2.3的GPU版本安装

一、总结

一句话总结：

安装tensorflow的gpu的版本的时候，要特别注意CUDA、cuDNN、tensorflow版本的一致性，在tensorflow官网可以查看对应版本关系

 

 

二、tensorflow2.3的GPU版本安装

博客对应课程的视频位置：

2、tensorflow2.3的GPU版本安装（一）-范仁义-读书编程笔记

https://www.fanrenyi.com/video/37/336

2、tensorflow2.3的GPU版本安装（二）-范仁义-读书编程笔记

https://www.fanrenyi.com/video/37/337

 

一、准备资料

需要显卡，还需要安装CUDA和cuDNN

1、显卡

比如我这台机器，显卡就是 NVIDIA GTX 1060，显存6GB

2、CUDA(Compute Unified Device Architecture)：统一计算架构

是显卡厂商NVIDIA推出的运算平台。
CUDA是一种由NVIDIA推出的通用并行计算架构，该架构使GPU能够解决复杂的计算问题。

3、CUDNN：NVIDIA cuDNN是用于深度神经网络的GPU加速库。

The NVIDIA CUDA® Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks.

4、CUDA和CUDNN的关系

CUDA看作是一个工作台，上面配有很多工具，如锤子、螺丝刀等。
cuDNN是基于CUDA的深度学习GPU加速库，有了它才能在GPU上完成深度学习的计算。

二、安装CUDA和cuDNN

特别注意：

当配置CUDA、cuDNN、tensorflow时，要确保这三者之间的版本对应一致

在tensorflow官网可以查看这些软件版本对应信息

https://tensorflow.google.cn/install/source_windows

1、安装cuda

各个版本位置

https://developer.nvidia.com/cuda-toolkit-archive

2、安装cuDNN

cuDNN各个版本的下载网址：

https://developer.nvidia.com/rdp/cudnn-archive#a-collapse51b

这个下载要注册，太太太麻烦，

所以我们可以直接复制链接地址，迅雷下载即可，不要点进去，点进去要注册麻烦

3、配置系统变量

 

 

4、验证是否安装成功

 

 

配置完成后，我们可以验证是否配置成功，主要使用CUDA内置的deviceQuery.exe 和 bandwithTest.exe：

首先win+R启动cmd，cd到安装目录下的 ...\extras\demo_suite,然后分别执行bandwidthTest.exe和deviceQuery.exe,

 

 

 

 

如果以上两步都返回了Result=PASS，那么就算成功啦。

三、安装tensorflow

如果安装了tensorflow-cpu版本，可以先把cpu版本先卸了

pip uninstall tensorflow-cpu

然后再安装gpu版本

pip install tensorflow-gpu

四、验证tensorflow是否成功用gpu加速

In [2]:

import tensorflow as tf print("GPU:",tf.test.is_gpu_available()) 

WARNING:tensorflow:From <ipython-input-2-ae62284c8fe5>:2: is_gpu_available (from tensorflow.python.framework.test_util) is deprecated and will be removed in a future version.
Instructions for updating:
Use `tf.config.list_physical_devices('GPU')` instead.
GPU: True

In [3]:

tf.config.list_physical_devices('GPU') 

Out[3]:

[PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')]

五、卷积神经网络中GUP和CPU性能对比

In [5]:

import  os os.environ['TF_CPP_MIN_LOG_LEVEL']='2' import tensorflow as tf from tensorflow.keras import layers, optimizers, datasets, Sequential # 按需分配GPU # from tensorflow.compat.v1 import ConfigProto # from tensorflow.compat.v1 import InteractiveSession # config = ConfigProto() # config.gpu_options.allow_growth = True # session = InteractiveSession(config=config) tf.random.set_seed(2345) # 卷积层 conv_layers = [ # 5 units of conv + max pooling # 5个单元 # unit 1 layers.Conv2D(64, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.Conv2D(64, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'), # unit 2 layers.Conv2D(128, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.Conv2D(128, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'), # unit 3 layers.Conv2D(256, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.Conv2D(256, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'), # unit 4 layers.Conv2D(512, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.Conv2D(512, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same'), # unit 5 layers.Conv2D(512, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.Conv2D(512, kernel_size=[3, 3], padding="same", activation=tf.nn.relu), layers.MaxPool2D(pool_size=[2, 2], strides=2, padding='same') ] # 归一化处理 def preprocess(x, y): # [0~1] x = tf.cast(x, dtype=tf.float32) / 255. y = tf.cast(y, dtype=tf.int32) return x,y # 加载数据 (x,y), (x_test, y_test) = datasets.cifar100.load_data() # 挤压 y = tf.squeeze(y, axis=1) y_test = tf.squeeze(y_test, axis=1) print(x.shape, y.shape, x_test.shape, y_test.shape) # 训练数据 train_db = tf.data.Dataset.from_tensor_slices((x,y)) train_db = train_db.shuffle(1000).map(preprocess).batch(128) # 测试数据 test_db = tf.data.Dataset.from_tensor_slices((x_test,y_test)) test_db = test_db.map(preprocess).batch(64) # 数据样例 sample = next(iter(train_db)) print('sample:', sample[0].shape, sample[1].shape, tf.reduce_min(sample[0]), tf.reduce_max(sample[0])) def main(): # 卷积层 # [b, 32, 32, 3] => [b, 1, 1, 512] conv_net = Sequential(conv_layers) # 全连接层 fc_net = Sequential([ layers.Dense(256, activation=tf.nn.relu), layers.Dense(128, activation=tf.nn.relu), # 100个分类，输出 layers.Dense(100, activation=None), ]) # 卷积层网络构建 conv_net.build(input_shape=[None, 32, 32, 3]) # 全连接网络构建 fc_net.build(input_shape=[None, 512]) # 优化函数 optimizer = optimizers.Adam(lr=1e-4) # [1, 2] + [3, 4] => [1, 2, 3, 4] # 所有的参数就是 卷积层的参数 和 全连接层的参数 variables = conv_net.trainable_variables + fc_net.trainable_variables for epoch in range(50): for step, (x,y) in enumerate(train_db): with tf.GradientTape() as tape: # 卷积层 # [b, 32, 32, 3] => [b, 1, 1, 512] out = conv_net(x) # reshape方便全连接层用 # flatten, => [b, 512] out = tf.reshape(out, [-1, 512]) # 全连接层 # [b, 512] => [b, 100] logits = fc_net(out) # [b] => [b, 100] y_onehot = tf.one_hot(y, depth=100) # compute loss：crossentropy loss = tf.losses.categorical_crossentropy(y_onehot, logits, from_logits=True) loss = tf.reduce_mean(loss) grads = tape.gradient(loss, variables) optimizer.apply_gradients(zip(grads, variables)) if step %100 == 0: print(epoch, step, 'loss:', float(loss)) total_num = 0 total_correct = 0 for x,y in test_db: out = conv_net(x) out = tf.reshape(out, [-1, 512]) logits = fc_net(out) prob = tf.nn.softmax(logits, axis=1) pred = tf.argmax(prob, axis=1) pred = tf.cast(pred, dtype=tf.int32) correct = tf.cast(tf.equal(pred, y), dtype=tf.int32) correct = tf.reduce_sum(correct) total_num += x.shape[0] total_correct += int(correct) acc = total_correct / total_num print(epoch, 'acc:', acc) if __name__ == '__main__': with tf.device("/cpu:0"): main() 

(50000, 32, 32, 3) (50000,) (10000, 32, 32, 3) (10000,)
sample: (128, 32, 32, 3) (128,) tf.Tensor(0.0, shape=(), dtype=float32) tf.Tensor(1.0, shape=(), dtype=float32)
0 0 loss: 4.605703353881836