Met the issue about TypeError when using lime_image.LimeImageExplainer()

def get_pil_transform(): 
    transf = transforms.Compose([
        transforms.Resize((224, 224)),
        # transforms.CenterCrop(224)
    ])    
    return transf

def batch_predict(images):
    model.eval()
    batch = torch.stack(tuple(eval_transforms(i) for i in images), dim=0)

    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    model.to(device)
    batch = batch.to(device)
    
    logits, _ = model(batch)
    probs = F.softmax(logits, dim=1)
    return probs.detach().cpu().numpy()

from lime import lime_image

img = get_image('../data_finetune/sample.jpg')
pil_transform = get_pil_transform()
transformed_img = pil_transform(img)
explainer = lime_image.LimeImageExplainer()
explanation = explainer.explain_instance(np.array(transformed_img), 
                                         batch_predict, # classification function
                                         top_labels=1, 
                                         hide_color=0, 
                                         num_samples=1000) # number of images that will be sent to classification function