如何将火炬张量转换为 Pytorch 中的 numpy arrays 列表？

Question

I have a torch tensor that looks like torch.Size([32, 3, 64, 64]).

I'm trying convert the tensor into something that can pass these assertions:

assert(type(images) == list)
assert(type(images[0]) == np.ndarray)
assert(len(images[0].shape) == 3)
assert(np.max(images[0]) > 10)
assert(np.min(images[0]) >= 0.0)

I'm currently doing this to convert the tensor:

# turn tensor into list of lists
imgs = imgs.tolist()

# iterate over list and turn each image into a numpy array with normalized values
for idx, img in enumerate(imgs):
  img = cv2.normalize(np.array(img), None,
  alpha = 0, beta = 255, norm_type = cv2.NORM_MINMAX )

and I get this error:

File "scripts/run_model.py", line 158, in get_inception_score
assert(type(images[0]) == np.ndarray)
AssertionError

How can I convert the tensor correctly so that type(images) is a list and type(images[0] is a np.ndarray)? Any help would be greatly appreciated. Thank you in advance.

Answer 1

Convert Pytorch tensor to numpy array first using tensor.numpy() and then convert it into a list using the built-in list() method.

images = torch.randn(32,3,64,64)
numpy_imgs = images.numpy()
list_imgs = list(numpy_imgs)
print(type(images))
print(type(numpy_imgs))
print(type(list_imgs))
print(type(list_imgs[0]))

<class 'torch.Tensor'>

<class 'numpy.ndarray'>

<class 'list'>

<class 'numpy.ndarray'>