我正在嘗試更多地了解計算機視覺模型，并且正在嘗試探索它們的工作原理。為了更好地理解如何解釋特征向量，我嘗試使用 Pytorch 來提取特征向量。下面是我從不同地方拼湊而成的代碼。import torchimport torch.nn as nnimport torchvision.models as modelsimport torchvision.transforms as transformsfrom torch.autograd import Variablefrom PIL import Imageimg=Image.open("Documents/01235.png")# Load the pretrained modelmodel = models.resnet18(pretrained=True)# Use the model object to select the desired layerlayer = model._modules.get('avgpool')# Set model to evaluation modemodel.eval()transforms = torchvision.transforms.Compose([        torchvision.transforms.Resize(256),        torchvision.transforms.CenterCrop(224),        torchvision.transforms.ToTensor(),        torchvision.transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]),    ])    def get_vector(image_name):    # Load the image with Pillow library    img = Image.open("Documents/Documents/Driven Data Competitions/Hateful Memes Identification/data/01235.png")    # Create a PyTorch Variable with the transformed image    t_img = transforms(img)    # Create a vector of zeros that will hold our feature vector    # The 'avgpool' layer has an output size of 512    my_embedding = torch.zeros(512)    # Define a function that will copy the output of a layer    def copy_data(m, i, o):        my_embedding.copy_(o.data)    # Attach that function to our selected layer    h = layer.register_forward_hook(copy_data)    # Run the model on our transformed image    model(t_img)    # Detach our copy function from the layer    h.remove()    # Return the feature vector    return my_embeddingpic_vector = get_vector(img)當我這樣做時，出現(xiàn)以下錯誤：RuntimeError: Expected 4-dimensional input for 4-dimensional weight [64, 3, 7, 7], but got 3-dimensional input of size [3, 224, 224] instead我確定這是一個基本錯誤，但我似乎無法弄清楚如何解決這個問題。我的印象是“totensor”轉(zhuǎn)換會使我的數(shù)據(jù)成為 4-d，但它似乎無法正常工作或者我誤解了它。感謝我可以用來了解更多信息的任何幫助或資源！