python常用输出形状方式

输出列表中张量形状：

import torch

features = [
    torch.randn(8, 64, 64, 64),
    torch.randn(8, 64, 32, 32),
    torch.randn(8, 64, 16, 16),
    torch.randn(8, 64, 8, 8)
]

for idx, feature in enumerate(features):
    print(f'feature {idx} shape: {feature.shape}')

取列表的最后一个元素predictions_class [-1] ：

predictions_class = [
    torch.Size([8, 64, 64, 64]),  
    torch.Size([8, 32, 32, 32]),  
    torch.Size([8, 16, 16, 16])   
]

last_prediction = predictions_class[-1]
print(last_prediction)  # 输出: torch.Size([8, 16, 16, 16])

输出张量形状：

print(tensor.shape)

字典由键(key)和值(values)成对表示：

input_shape = {
            'res2': ShapeInfo(channels=64, stride=1),
            'res3': ShapeInfo(channels=64, stride=2),
            'res4': ShapeInfo(channels=64, stride=3),
            'res5': ShapeInfo(channels=64, stride=4)
        }

按stride对上述四组排序，输出字典内容： 

sorted_input_shape = sorted(input_shape.items(), key=lambda x: x[1].stride)
self.in_features = [k for k, v in sorted_input_shape] 
feature_channels = [v.channels for k, v in sorted_input_shape]

print("In features:", self.in_features)
print("Feature channels:", feature_channels)

已知输入的张量tensor，获取对应的高度和宽度：

tensor = torch.randn(8, 64, 64, 64)
_, _, height, width = tensor.shape

print(f"Height: {height}")
print(f"Width: {width}")

将张量沿着dim=1的维度进行划分：

x.narrow(1, start, length)