devbox
盐析白兔
这个屌丝很懒，什么也没留下！
关注作者
热门标签
热门文章
当前位置:   article > 正文

在2023年，在Ubuntu 18中 安装tensorflow 1.14.0+cuda 10.0/10.2 + 对应版本的cudnn 7.6.5 +GTX 960/2080ti/QuadroP600_ubuntu 安装cuda和tensorflow1.14

作者：盐析白兔 | 2024-02-22 05:01:52

赞

踩

ubuntu 安装cuda和tensorflow1.14

也是为了复现最原始的 Attention is all you need 实验过程；

1. cuda10.0 

一， 安装cuda10.0 和 对应的 cudnn7.6.5

下载cuda 10.0 for ubuntu18.04

CUDA Toolkit 10.0 Download | NVIDIA Developer

安装cuda：
 
预备工作

创建blacklist***.conf具体如下，加入两行内容

  1.  
  2.  
  3. sudo vim /etc/modprobe.d/blacklist-nouveau.conf
  4. 内容：
  5. blacklist nouveau
  6. options nouveau modeset=0
  7.  
  8. sudo update-initramfs -u
  9.  
  10.  
  11.  
  12.  
  13.

安装cuda 10.0：

1. 手动下载 cuda 10.0 for  ubuntu18.04，并安装：

  1.  
  2. sudo apt-get install linux-headers-$(uname -r)   \
  3. && sudo apt-key del 7fa2af80  \
  4. && sudo dpkg -i cuda-repo-ubuntu1804-10-0-local-10.0.130-410.48_1.0-1_amd64.deb \
  5. && sudo apt-key add /var/cuda-repo-10-0-local-10.0.130-410.48/7fa2af80.pub \
  6. && sudo apt-get update \
  7. && sudo apt-get -y install cuda


注册nvidia开发者，并下载对应 ***cudnn***.deb,安装命令： 

  1. #注意安装顺序：
  2.  
  3.  
  4. sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.0_amd64.deb && \
  5. sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.0_amd64.deb && \
  6. sudo dpkg -i libcudnn7-doc_7.6.5.32-1+cuda10.0_amd64.deb

二，安装python3 环境

预先安装合适版本的python3   module：

  1. sudo apt install python3.7*
  2.  
  3. sudo rm /usr/bin/python3
  4.  
  5. sudo ln /usr/bin/python3.7 /usr/bin/python3
  6.  
  7. sudo apt install python3-pip
  8.  
  9. sudo pip3 install ipython
  10.  
  11. sudo pip3 install Cython
  12.  
  13. sudo apt-get install -y glibc-doc manpages-posix-dev
  14.  
  15. sudo pip3 install numpy==1.15.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/
  16.  
  17. sudo pip3 install pkgconfig==1.4.0
  18.  
  19. cd /usr/lib/python3/dist-packages
  20. sudo ln -s apt_pkg.cpython-36m-x86_64-linux-gnu.so apt_pkg.so
  21.  
  22. # install HDF5:具体方法在文末
  23.  
  24. sudo pip3 install h5py==2.10.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/
  25.  
  26. sudo pip3 install setuptools==57.5.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/

三，安装tensorflow 1.14.0
 sudo pip3 install tensorflow-gpu==1.14.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/

四，下载，编译，安装hdf5

HDF5 1.10.10

  1.  
  2.  
  3. sudo apt-get update
  4. sudo apt-get install build-essential
  5. sudo apt-get build-dep hdf5
  6.  
  7. mkdir ~/Software
  8. cd ~/Software
  9. wget https://support.hdfgroup.org/ftp/HDF5/releases/hdf5-1.10/hdf5-1.10.10/src/hdf5-1.10.10.tar.gz
  10.  
  11. tar -xf hdf5-1.8.10.tar.gz
  12. cd hdf5-1.8.10/
  13. ./configure
  14. make -j9
  15. sudo make install

五，测试tensorflow的安装效果

demo 测试源码   hello_tf114.py

  1. import os
  2. os.environ['TF_CPP_MIN_LOG_LEVEL']='2'
  3.  
  4. def tf114_demo():
  5.         a = 3
  6.         b = 4
  7.         c = a + b
  8.         print("a + b in py =",c)
  9.  
  10.         a_t = tf.constant(3)
  11.         b_t = tf.constant(4)
  12.         c_t = a_t + b_t
  13.         print("TensorFlow add a_t + b_t =", c_t)
  14.  
  15.         with tf.Session() as sess:
  16.                 c_t_value = sess.run(c_t)
  17.                 print("c_t_value= ", c_t_value)
  18.  
  19.         return None
  20.  
  21. if __name__ == "__main__":
  22.         tf114_demo()

运行结果：

2. cuda10.2 + rtx2080ti

一，尝试使用cuda10.2来支持tensorflow1.14.0

rtx 2080ti 是2018年5月发布，cuda 10.2 首版是2019年发布的，故可以推断，cuda 10.2也支持2080ti，并可以在 2080ti上 运行  tensorflow 1.14.0；待更深入的网络训练的测试

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

实测发现，在ubuntu18.04中，使用cuda10.2并不能直接支持tensorflow 1.14.0，因为tensorflow 1.14.0会默认 load cuda相关lib的10.0版本，例如 libcudart.so.10.0   libcublas.so.10.0 等；

如果安装cuda10.2后不做任何改动，将出现如下错误：

所以，要尝试用10.2的so充当 10.0 的 so文件：

  1. cd /usr/lib/x86_64-linux-gnu \
  2. && sudo ln -s libcublas.so.10.2.2.89 libcublas.so.10.0 \
  3. && sudo ln -s libcublasLt.so.10.2.2.89 libcublasLt.so.10.0 
  4.  
  5. export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-10.2/targets/x86_64-linux/lib/
  6.  
  7. cd  /usr/local/cuda-10.2/targets/x86_64-linux/lib/ \
  8. && sudo ln -s   libcufft.so.10.1.2.89         libcufft.so.10.0            \
  9. && sudo ln -s   libcufftw.so.10.1.2.89        libcufftw.so.10.0        \
  10. && sudo ln -s   libcurand.so.10.1.2.89        libcurand.so.10.0        \
  11. && sudo ln -s   libcusolverMg.so.10.3.0.89    libcusolverMg.so.10.0    \
  12. && sudo ln -s   libcusolver.so.10.3.0.89      libcusolver.so.10.0      \
  13. && sudo ln -s   libcusparse.so.10.3.1.89      libcusparse.so.10.0      \
  14. && sudo ln -s   libnppc.so.10.2.1.89          libnppc.so.10.0          \
  15. && sudo ln -s   libnppial.so.10.2.1.89        libnppial.so.10.0        \
  16. && sudo ln -s   libnppicc.so.10.2.1.89        libnppicc.so.10.0        \
  17. && sudo ln -s   libnppicom.so.10.2.1.89       libnppicom.so.10.0       \
  18. && sudo ln -s   libnppidei.so.10.2.1.89       libnppidei.so.10.0       \
  19. && sudo ln -s   libnppif.so.10.2.1.89         libnppif.so.10.0         \
  20. && sudo ln -s   libnppig.so.10.2.1.89         libnppig.so.10.0         \
  21. && sudo ln -s   libnppim.so.10.2.1.89         libnppim.so.10.0         \
  22. && sudo ln -s   libnppist.so.10.2.1.89        libnppist.so.10.0        \
  23. && sudo ln -s   libnppisu.so.10.2.1.89        libnppisu.so.10.0        \
  24. && sudo ln -s   libnppitc.so.10.2.1.89        libnppitc.so.10.0        \
  25. && sudo ln -s   libnpps.so.10.2.1.89          libnpps.so.10.0          \
  26. && sudo ln -s   libnvgraph.so.10.2.89         libnvgraph.so.10.0       \
  27. && sudo ln -s   libnvjpeg.so.10.3.1.89        libnvjpeg.so.10.0        \
  28. && sudo ln -s  libcudart.so.10.2.89           libcudart.so.10.0            \
  29. && sudo ln -s  libaccinj64.so.10.2.89         libaccinj64.so.10.0          \
  30. && sudo ln -s  libcuinj64.so.10.2.89          libcuinj64.so.10.0           \
  31. && sudo ln -s  libcupti.so.10.2.75            libcupti.so.10.0             \
  32. && sudo ln -s  libnvrtc-builtins.so.10.2.89   libnvrtc-builtins.so.10.0    \
  33. && sudo ln -s  libnvrtc.so.10.2.89            libnvrtc.so.10.0    
  34.  
  35.  
  36.  
  37. #别忘了这个：
  38. export  LD_LIBRARY_PATH=/usr/local/cuda-10.2/targets/x86_64-linux/lib/

export  LD_LIBRARY_PATH=/usr/local/cuda-10.2/targets/x86_64-linux/lib/

浅测可以：

cuda10.0 应该是调用不动2080TI卡，因为发布时间的先后；而 cuda10.2发布的时间晚于2080ti的发布时间，所以是可以支持到的；

二，深入的测试 demo_01

bidirectional_rnn.py

  1. """ Bi-directional Recurrent Neural Network.
  3. A Bi-directional Recurrent Neural Network (LSTM) implementation example using
  4. TensorFlow library. This example is using the MNIST database of handwritten
  5. digits (http://yann.lecun.com/exdb/mnist/)
  7. Links:
  8.     [Long Short Term Memory](http://deeplearning.cs.cmu.edu/pdfs/Hochreiter97_lstm.pdf)
  9.     [MNIST Dataset](http://yann.lecun.com/exdb/mnist/).
  11. Author: Aymeric Damien
  12. Project: https://github.com/aymericdamien/TensorFlow-Examples/
  13. """
  14.  
  15. from __future__ import print_function
  16.  
  17. import tensorflow as tf
  18. from tensorflow.contrib import rnn
  19. import numpy as np
  20.  
  21. # Import MNIST data
  22. from tensorflow.examples.tutorials.mnist import input_data
  23. mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
  24.  
  25. '''
  26. To classify images using a bidirectional recurrent neural network, we consider
  27. every image row as a sequence of pixels. Because MNIST image shape is 28*28px,
  28. we will then handle 28 sequences of 28 steps for every sample.
  29. '''
  30.  
  31. # Training Parameters
  32. learning_rate = 0.001
  33. training_steps = 10000
  34. batch_size = 128
  35. display_step = 200
  36.  
  37. # Network Parameters
  38. num_input = 28 # MNIST data input (img shape: 28*28)
  39. timesteps = 28 # timesteps
  40. num_hidden = 128 # hidden layer num of features
  41. num_classes = 10 # MNIST total classes (0-9 digits)
  42.  
  43. # tf Graph input
  44. X = tf.placeholder("float", [None, timesteps, num_input])
  45. Y = tf.placeholder("float", [None, num_classes])
  46.  
  47. # Define weights
  48. weights = {
  49.     # Hidden layer weights => 2*n_hidden because of forward + backward cells
  50.     'out': tf.Variable(tf.random_normal([2*num_hidden, num_classes]))
  51. }
  52. biases = {
  53.     'out': tf.Variable(tf.random_normal([num_classes]))
  54. }
  55.  
  56.  
  57. def BiRNN(x, weights, biases):
  58.  
  59.     # Prepare data shape to match `rnn` function requirements
  60.     # Current data input shape: (batch_size, timesteps, n_input)
  61.     # Required shape: 'timesteps' tensors list of shape (batch_size, num_input)
  62.  
  63.     # Unstack to get a list of 'timesteps' tensors of shape (batch_size, num_input)
  64.     x = tf.unstack(x, timesteps, 1)
  65.  
  66.     # Define lstm cells with tensorflow
  67.     # Forward direction cell
  68.     lstm_fw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)
  69.     # Backward direction cell
  70.     lstm_bw_cell = rnn.BasicLSTMCell(num_hidden, forget_bias=1.0)
  71.  
  72.     # Get lstm cell output
  73.     try:
  74.         outputs, _, _ = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
  75.                                               dtype=tf.float32)
  76.     except Exception: # Old TensorFlow version only returns outputs not states
  77.         outputs = rnn.static_bidirectional_rnn(lstm_fw_cell, lstm_bw_cell, x,
  78.                                         dtype=tf.float32)
  79.  
  80.     # Linear activation, using rnn inner loop last output
  81.     return tf.matmul(outputs[-1], weights['out']) + biases['out']
  82.  
  83. logits = BiRNN(X, weights, biases)
  84. prediction = tf.nn.softmax(logits)
  85.  
  86. # Define loss and optimizer
  87. loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
  88.     logits=logits, labels=Y))
  89. optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
  90. train_op = optimizer.minimize(loss_op)
  91.  
  92. # Evaluate model (with test logits, for dropout to be disabled)
  93. correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
  94. accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
  95.  
  96. # Initialize the variables (i.e. assign their default value)
  97. init = tf.global_variables_initializer()
  98.  
  99. # Start training
  100. with tf.Session() as sess:
  101.  
  102.     # Run the initializer
  103.     sess.run(init)
  104.  
  105.     for step in range(1, training_steps+1):
  106.         batch_x, batch_y = mnist.train.next_batch(batch_size)
  107.         # Reshape data to get 28 seq of 28 elements
  108.         batch_x = batch_x.reshape((batch_size, timesteps, num_input))
  109.         # Run optimization op (backprop)
  110.         sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})
  111.         if step % display_step == 0 or step == 1:
  112.             # Calculate batch loss and accuracy
  113.             loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,
  114.                                                                  Y: batch_y})
  115.             print("Step " + str(step) + ", Minibatch Loss= " + \
  116.                   "{:.4f}".format(loss) + ", Training Accuracy= " + \
  117.                   "{:.3f}".format(acc))
  118.  
  119.     print("Optimization Finished!")
  120.  
  121.     # Calculate accuracy for 128 mnist test images
  122.     test_len = 128
  123.     test_data = mnist.test.images[:test_len].reshape((-1, timesteps, num_input))
  124.     test_label = mnist.test.labels[:test_len]
  125.     print("Testing Accuracy:", \
  126.         sess.run(accuracy, feed_dict={X: test_data, Y: test_label}))

$ python3  bidirectional_rnn.py


 

训练效果：

 nvidia-smi  gpu 使用效果：

三，深入的测试 demo_02

 convolutional_network_raw_deviceInfo.py

  1. """ Convolutional Neural Network.
  3. Build and train a convolutional neural network with TensorFlow.
  4. This example is using the MNIST database of handwritten digits
  5. (http://yann.lecun.com/exdb/mnist/)
  7. Author: Aymeric Damien
  8. Project: https://github.com/aymericdamien/TensorFlow-Examples/
  9. """
  10.  
  11. from __future__ import division, print_function, absolute_import
  12.  
  13. import tensorflow as tf
  14.  
  15. # Import MNIST data
  16. from tensorflow.examples.tutorials.mnist import input_data
  17. mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)
  18.  
  19. # Training Parameters
  20. learning_rate = 0.001
  21. num_steps = 200
  22. batch_size = 128
  23. display_step = 10
  24.  
  25. # Network Parameters
  26. num_input = 784 # MNIST data input (img shape: 28*28)
  27. num_classes = 10 # MNIST total classes (0-9 digits)
  28. dropout = 0.75 # Dropout, probability to keep units
  29.  
  30. # tf Graph input
  31. X = tf.placeholder(tf.float32, [None, num_input])
  32. Y = tf.placeholder(tf.float32, [None, num_classes])
  33. keep_prob = tf.placeholder(tf.float32) # dropout (keep probability)
  34.  
  35.  
  36. # Create some wrappers for simplicity
  37. def conv2d(x, W, b, strides=1):
  38.     # Conv2D wrapper, with bias and relu activation
  39.     x = tf.nn.conv2d(x, W, strides=[1, strides, strides, 1], padding='SAME')
  40.     x = tf.nn.bias_add(x, b)
  41.     return tf.nn.relu(x)
  42.  
  43.  
  44. def maxpool2d(x, k=2):
  45.     # MaxPool2D wrapper
  46.     return tf.nn.max_pool(x, ksize=[1, k, k, 1], strides=[1, k, k, 1],
  47.                           padding='SAME')
  48.  
  49.  
  50. # Create model
  51. def conv_net(x, weights, biases, dropout):
  52.     # MNIST data input is a 1-D vector of 784 features (28*28 pixels)
  53.     # Reshape to match picture format [Height x Width x Channel]
  54.     # Tensor input become 4-D: [Batch Size, Height, Width, Channel]
  55.     x = tf.reshape(x, shape=[-1, 28, 28, 1])
  56.  
  57.     # Convolution Layer
  58.     conv1 = conv2d(x, weights['wc1'], biases['bc1'])
  59.     # Max Pooling (down-sampling)
  60.     conv1 = maxpool2d(conv1, k=2)
  61.  
  62.     # Convolution Layer
  63.     conv2 = conv2d(conv1, weights['wc2'], biases['bc2'])
  64.     # Max Pooling (down-sampling)
  65.     conv2 = maxpool2d(conv2, k=2)
  66.  
  67.     # Fully connected layer
  68.     # Reshape conv2 output to fit fully connected layer input
  69.     fc1 = tf.reshape(conv2, [-1, weights['wd1'].get_shape().as_list()[0]])
  70.     fc1 = tf.add(tf.matmul(fc1, weights['wd1']), biases['bd1'])
  71.     fc1 = tf.nn.relu(fc1)
  72.     # Apply Dropout
  73.     fc1 = tf.nn.dropout(fc1, dropout)
  74.  
  75.     # Output, class prediction
  76.     out = tf.add(tf.matmul(fc1, weights['out']), biases['out'])
  77.     return out
  78.  
  79. # Store layers weight & bias
  80. weights = {
  81.     # 5x5 conv, 1 input, 32 outputs
  82.     'wc1': tf.Variable(tf.random_normal([5, 5, 1, 32])),
  83.     # 5x5 conv, 32 inputs, 64 outputs
  84.     'wc2': tf.Variable(tf.random_normal([5, 5, 32, 64])),
  85.     # fully connected, 7*7*64 inputs, 1024 outputs
  86.     'wd1': tf.Variable(tf.random_normal([7*7*64, 1024])),
  87.     # 1024 inputs, 10 outputs (class prediction)
  88.     'out': tf.Variable(tf.random_normal([1024, num_classes]))
  89. }
  90.  
  91. biases = {
  92.     'bc1': tf.Variable(tf.random_normal([32])),
  93.     'bc2': tf.Variable(tf.random_normal([64])),
  94.     'bd1': tf.Variable(tf.random_normal([1024])),
  95.     'out': tf.Variable(tf.random_normal([num_classes]))
  96. }
  97.  
  98. # Construct model
  99. logits = conv_net(X, weights, biases, keep_prob)
  100. prediction = tf.nn.softmax(logits)
  101.  
  102. # Define loss and optimizer
  103. loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
  104.     logits=logits, labels=Y))
  105. optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
  106. train_op = optimizer.minimize(loss_op)
  107.  
  108.  
  109. # Evaluate model
  110. correct_pred = tf.equal(tf.argmax(prediction, 1), tf.argmax(Y, 1))
  111. accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))
  112.  
  113. # Initialize the variables (i.e. assign their default value)
  114. init = tf.global_variables_initializer()
  115.  
  116. # Start training
  117. with tf.Session(config=tf.ConfigProto(allow_soft_placement=True, log_device_placement=True)) as sess:
  118.  
  119.     # Run the initializer
  120.     sess.run(init)
  121.  
  122.     for step in range(1, num_steps+1):
  123.         batch_x, batch_y = mnist.train.next_batch(batch_size)
  124.         # Run optimization op (backprop)
  125.         sess.run(train_op, feed_dict={X: batch_x, Y: batch_y, keep_prob: 0.8})
  126.         if step % display_step == 0 or step == 1:
  127.             # Calculate batch loss and accuracy
  128.             loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,
  129.                                                                  Y: batch_y,
  130.                                                                  keep_prob: 1.0})
  131.             print("Step " + str(step) + ", Minibatch Loss= " + \
  132.                   "{:.4f}".format(loss) + ", Training Accuracy= " + \
  133.                   "{:.3f}".format(acc))
  134.  
  135.     print("Optimization Finished!")
  136.  
  137.     # Calculate accuracy for 256 MNIST test images
  138.     print("Testing Accuracy:", \
  139.         sess.run(accuracy, feed_dict={X: mnist.test.images[:256],
  140.                                       Y: mnist.test.labels[:256],
  141.                                       keep_prob: 1.0}))

$ python3    convolutional_network_raw_deviceInfo.py

训练效果：

 $nvidia-smi

————————————————————————————————————————

  1. sudo apt install python3-pip \
  2. && sudo pip3 install Cython \
  3. && sudo pip3 install ipython \
  4. && sudo apt-get install -y glibc-doc manpages-posix-dev \
  5. && sudo pip3 install numpy==1.15.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/ \
  6. && sudo pip3 install pkgconfig==1.4.0 \
  7. && cd /usr/lib/python3/dist-packages \
  8. && sudo ln -s apt_pkg.cpython-36m-x86_64-linux-gnu.so apt_pkg.so
  9.  
  10.  
  11.  
  12.  
  13. sudo apt-get update \
  14. && sudo apt-get install build-essential \
  15. && sudo apt-get build-dep hdf5 \
  16. && mkdir ~/Software \
  17. && cd ~/Software \
  18. && wget https://support.hdfgroup.org/ftp/HDF5/releases/hdf5-1.10/hdf5-1.10.10/src/hdf5-1.10.10.tar.gz \
  19. && tar -xf hdf5-1.10.10.tar.gz \
  20. && cd hdf5-1.10.10/ \
  21. && ./configure \
  22. && make -j9 \
  23. && sudo make install
  24.  
  25.  
  26.  
  27. sudo pip3 install h5py==2.10.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/ \
  28. && sudo pip3 install setuptools==57.5.0 -i https://pypi.tuna.tsinghua.edu.cn/simple/
  29.  
  30. cd /usr/lib/x86_64-linux-gnu \
  31. && sudo ln -s libcublas.so.10.2.2.89 libcublas.so.10.0 \
  32. && sudo ln -s libcublasLt.so.10.2.2.89 libcublasLt.so.10.0 
  33.  
  34.  
  35.  
  36.  
  37.  
  38.  
  39.  
  40.  
  41.  
  42.  
  43.  
  44. cd  /usr/local/cuda-10.2/targets/x86_64-linux/lib/ \
  45. && sudo ln -s   libcufft.so.10.1.2.89         libcufft.so.10.0            \
  46. && sudo ln -s   libcufftw.so.10.1.2.89        libcufftw.so.10.0        \
  47. && sudo ln -s   libcurand.so.10.1.2.89        libcurand.so.10.0        \
  48. && sudo ln -s   libcusolverMg.so.10.3.0.89    libcusolverMg.so.10.0    \
  49. && sudo ln -s   libcusolver.so.10.3.0.89      libcusolver.so.10.0      \
  50. && sudo ln -s   libcusparse.so.10.3.1.89      libcusparse.so.10.0      \
  51. && sudo ln -s   libnppc.so.10.2.1.89          libnppc.so.10.0          \
  52. && sudo ln -s   libnppial.so.10.2.1.89        libnppial.so.10.0        \
  53. && sudo ln -s   libnppicc.so.10.2.1.89        libnppicc.so.10.0        \
  54. && sudo ln -s   libnppicom.so.10.2.1.89       libnppicom.so.10.0       \
  55. && sudo ln -s   libnppidei.so.10.2.1.89       libnppidei.so.10.0       \
  56. && sudo ln -s   libnppif.so.10.2.1.89         libnppif.so.10.0         \
  57. && sudo ln -s   libnppig.so.10.2.1.89         libnppig.so.10.0         \
  58. && sudo ln -s   libnppim.so.10.2.1.89         libnppim.so.10.0         \
  59. && sudo ln -s   libnppist.so.10.2.1.89        libnppist.so.10.0        \
  60. && sudo ln -s   libnppisu.so.10.2.1.89        libnppisu.so.10.0        \
  61. && sudo ln -s   libnppitc.so.10.2.1.89        libnppitc.so.10.0        \
  62. && sudo ln -s   libnpps.so.10.2.1.89          libnpps.so.10.0          \
  63. && sudo ln -s   libnvgraph.so.10.2.89         libnvgraph.so.10.0       \
  64. && sudo ln -s   libnvjpeg.so.10.3.1.89        libnvjpeg.so.10.0        \
  65. && sudo ln -s  libcudart.so.10.2.89           libcudart.so.10.0            \
  66. && sudo ln -s  libaccinj64.so.10.2.89         libaccinj64.so.10.0          \
  67. && sudo ln -s  libcuinj64.so.10.2.89          libcuinj64.so.10.0           \
  68. && sudo ln -s  libcupti.so.10.2.75            libcupti.so.10.0             \
  69. && sudo ln -s  libnvrtc-builtins.so.10.2.89   libnvrtc-builtins.so.10.0    \
  70. && sudo ln -s  libnvrtc.so.10.2.89            libnvrtc.so.10.0
  71.  
  72.  
  73.  
  74. export  LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-10.2/targets/x86_64-linux/lib/

测试Quadro  P600 4GB显卡也可以， cuda 10.2 + cudnn 7.6.5 + tensorflow 1.14.0

 成功训练：

 

 

声明：本文内容由网友自发贡献，不代表【wpsshop博客】立场，版权归原作者所有，本站不承担相应法律责任。如您发现有侵权的内容，请联系我们。转载请注明出处：https://www.wpsshop.cn/w/盐析白兔/article/detail/128599
推荐阅读
相关标签

Copyright © 2003-2013 www.wpsshop.cn 版权所有，并保留所有权利。

  

闽ICP备14008679号