TypeError: can’t convert CUDA tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first

Question

I am using a modified predict.py for testing a pruned SqueezeNet Model

[phung@archlinux SqueezeNet-Pruning]$ python predict.py --image 3_100.jpg --model model_prunned --num_class 2
prediction in progress
Traceback (most recent call last):
File “predict.py”, line 66, in
prediction = predict_image(imagepath)
File “predict.py”, line 52, in predict_image
index = output.data.numpy().argmax()
TypeError: can’t convert CUDA tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
[phung@archlinux SqueezeNet-Pruning]$

I understand that numpy does not support GPU yet.

How shall I modify the code to get away from this error without invoking tensor copy data operation, tensor.cpu() ?

Answer 1

Change

index = output.data.numpy().argmax()

to

index = output.cpu().data.numpy().argmax()

This means data is first moved to cpu and then converted to numpy array.

Answer 2

I found out that I can just use

output.argmax()

Answer 3

You can use torch.max function like follows:

value, index = torch.max(output,1)