ScoreCAM got an all NaN result

[Yana-Hangabina]

Hi, dear developer, I ran my code and had this problem, while GradCAM, GCpp, CAM are all fine. Plz help me check this😭

import os
from PIL import Image
import numpy as np
import torch
import torch.nn.functional as F
from torch.autograd import Variable
from torchvision import models
from torchvision.transforms import transforms
from torchvision.transforms.functional import normalize, resize, to_pil_image
from torchcam.methods import ScoreCAM

H, W = 224, 224

transform = transforms.Compose(
        [
                #transforms.Resize((H, W)),
                transforms.ToTensor(),
                transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
        ])

def load_image(path, preprocess=False):
    img = Image.open(path).convert('RGB').resize((H, W))
    if preprocess:
        img = transform(img)
        img = img.unsqueeze(0)
    return img

def build_model(arch):
    model = models.__dict__[arch](weights='DEFAULT')
    # model = model.eval()
    model = model.cuda()
    return model

def dg(model, img_path, extractor):
    img = load_image(img_path, preprocess=True).cuda()
    
    with torch.no_grad(): out = model(img)
    cls = out.squeeze(0).argmax().item()
    
    cams = extractor(cls)
    cam = cams[0]
    ...
    return ...
    
model = build_model(arch='resnet50')

cams = {
    'ScoreCAM': ScoreCAM(model, 'layer4')
}

img_dir = './ImageNet2012_val'
img_paths = [os.path.join(img_dir, f) for f in os.listdir(img_dir)]
img_num = len(img_paths)

for img_path in img_paths:
    for cam, cam_extractor in cams.items():
        dd = dg(model, img_path, cam_extractor)
        ........

And this is the result value of cam variable when I debug, any random image inputed result the same.

tensor([[[nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan],
         [nan, nan, nan, nan, nan, nan, nan]]], device='cuda:0')

Environment: torch-cam 0.3.2

[Yana-Hangabina]

Oh I upgraded to he latest 0.4.0dev，ScoreCAM is fine now.
However when I delete with torch.no_grad(): and change to run GradCAM and other gradient-based methods, it will get an error CUDA error: CUBLAS_STATUS_INVALID_VALUE when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc).
If I use CPU to run it, then gradient-based methods will be fine but activation based methods will raise the error RuntimeError: 'cannot register a hook on a tensor that doesn't require gradient'.

[frgfm]

Glad it works now 😁

About your error:

• what is the full traceback of the error? (this is a CUDA error, but it has been triggered by one python line somewhere)
• about your error on CPU, could you provide a minimal snippet to reproduce this please? 🙏