Can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first

Question

I am trying to show results of GAN network on some specified epochs. The function for printing the current result was used previously with TF. I need to change in to pytorch.

def show_result(G_net, z_, num_epoch, show=False, save=False, path='result.png'):


  #test_images = sess.run(G_z, {z: z_, drop_out: 0.0})
  test_images = G_net(z_)

  size_figure_grid = 5
  fig, ax = plt.subplots(size_figure_grid, size_figure_grid, figsize=(5, 5))

  for i, j in itertools.product(range(size_figure_grid), range(size_figure_grid)):
     ax[i, j].get_xaxis().set_visible(False)
     ax[i, j].get_yaxis().set_visible(False)

  for k in range(5*5):
     i = k // 5
     j = k % 5
     ax[i, j].cla()
     ax[i, j].imshow(np.reshape(test_images[k], (28, 28)), cmap='gray')

  label = 'Epoch {0}'.format(num_epoch)
  fig.text(0.5, 0.04, label, ha='center')

  plt.savefig(name)
  file = drive.CreateFile({'title': label, "parents": [{"kind": "https://drive.google.com/drive/u/0/folders/", "id": folder_id}]})
  file.SetContentFile(name)
  file.Upload()

  if num_epoch == 10 or num_epoch == 20 or num_epoch == 50 or num_epoch == 100:
     plt.show()
    
  plt.close()

The results I need to obtain looks like that: result img

I am getting this error, however I am not sure what I did incorrectly

Can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first

Answer 1

I am assuming G_net is your network. It looks like you store the network in the GPU, hence the returned result test_images will also be in GPU. You will need to move it to cpu and convert to numpy:

#test_images = G_net(z_)
test_images = G_net(z_).detach().cpu().numpy()

This will detach the tensor from the graph, move to cpu and then convert to numpy.