pytorch相关
pytorch、cuda、cudnn对应关系
:
https://blog.csdn.net/caiguanhong/article/details/112184290
pip 安装:
pip install torch==1.6.0+cu101 torchvision==0.7.0+cu101 -f https://download.pytorch.org/whl/torch_stable.html
注意:pip 或者conda 按照以上方法安装的torch,可能还是报错或gpu跑不起来。
建议
去下面链接里找对应正确的torch 和torchvision 的 whl包安装(torch的版本需要对应的cuda,python,系统等)
torch/torchvision 各个版本链接(重要!!!!):
https://download.pytorch.org/whl/torch_stable.html
检查pytorch gpu是否安装好:
import torchvision
print(torch.__version__)
print(torchvision.__version__)
print(torch.cuda.is_available())
tensorflow相关
虚拟环境安装tensorflow(通过指定镜像源)
pip3 install -i
Simple Index
–upgrade tensorflow-gpu==2.4.1
# 当遇到“Could not fetch URL
Links for tensorflow-gpu
: There was a problem confirming the ssl certificate: [SSL: CERTIFICATE_VERIFY_FAILED] certificate verify failed (_ssl.c:833) – skippin”的报错时 :
pip install tensorflow-gpu==1.13.2 -i
Simple Index
–trusted-host
pypi.tuna.tsinghua.edu.cn
安装对应版本的tensorflow
:(不同镜像源的速度不同,可以选用最快的一个)
pip install tensorflow-gpu==2.3.0 -i
Simple Index
测试安装的 tensorflow 是否可用
测试一:
import tensorflow as tf
print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))
测试二:
print(tf.test.is_gpu_available())
测试三:
import tensorflow as tf
tf.compat.v1.disable_eager_execution()
config = tf.compat.v1.ConfigProto(allow_soft_placement=True)
sess = tf.compat.v1.Session(config=config)
with tf.device('/gpu:0'):
a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
c = tf.matmul(a, b)
print(sess.run(c))