RuntimeError: Cannot call numpy() on Tensor that requires grad

[woctezuma]

When running the first usage example on a machine without GPU (on Heroku), I have encountered an error message.

This is the usage example:

import torch
import clip
from PIL import Image

device = "cuda" if torch.cuda.is_available() else "cpu"
model, preprocess = clip.load("ViT-B/32", device=device)

image = preprocess(Image.open("CLIP.png")).unsqueeze(0).to(device)
text = clip.tokenize(["a diagram", "a dog", "a cat"]).to(device)

with torch.no_grad():
    image_features = model.encode_image(image)
    text_features = model.encode_text(text)
    
    logits_per_image, logits_per_text = model(image, text)
    probs = logits_per_image.softmax(dim=-1).cpu().numpy()

print("Label probs:", probs)  # prints: [[0.9927937  0.00421068 0.00299572]]

This is the error message:

RuntimeError: Cannot call numpy() on Tensor that requires grad 

I have followed the recommendation displayed along the error message, and changed this line from:

    probs = logits_per_image.softmax(dim=-1).cpu().numpy()

to:

    probs = logits_per_image.softmax(dim=-1).cpu().detach().numpy()

I have no idea whether the issue is specific to my machine, but I might as well mention it here.

[jongwook]

That's strange, since torch.no_grad() should have disabled any gradient computation. Although detach() did the trick for you, there's a chance that your code is using unnecessary memory/compute to retain gradient information.

Could you double-check if the numpy() line is under the with torch.no_grad(): block, and if it's the case (and if you're comfortable doing this) could you paste the output of the following script:

curl -s https://raw.githubusercontent.com/pytorch/pytorch/master/torch/utils/collect_env.py | python

to get the version information of your system and relevant libraries?

[woctezuma]

Could you double-check if the numpy() line is under the with torch.no_grad(): block

You are right. I did not write the .numpy() call under the with torch.no_grad() block in my file. Sorry, my bad!

This is what I had written in my copy-paste of your usage example:

    with torch.no_grad():
        image_features = model.encode_image(image)
        text_features = model.encode_text(text)
    
    logits_per_image, logits_per_text = model(image, text)
    probs = logits_per_image.softmax(dim=-1).cpu().detach().numpy()

I did not realize that I had changed the indentation.