Runtimeerror: numpy is not available solution

Problem Description:

Today, the old computer crashed (cried) and the new computer got started. I couldn’t wait to install the pytorch and check the operation of the previous project. As a result, there was an error running the model training

the key sentence is the sentence loading the model training data

for i, data in enumerate(train_loader):

The bottom layer of this statement is applied to the operation of converting the tensor variable to numpy, that is

import numpy
import torch
test_torch = torch.rand(100,1,28,28)
print(test_torch.numpy())
# will produce the same error

Problem root

The direct correlation between pytoch and tensorflow is ignored
if there is only torch but no tensorflow in the environment, the tensor variable used by torch will lose the tensor related operation function, such as torch. Numpy()

Solution:

Install tensorflow in the basic environment, whether CPU version or GPU version
in Anaconda environment, you can directly

conda install tensorflow

Otherwise pip

pip install --ignore-installed --upgrade tensorflow-gpu