[人工智能]查看模型的内部信息

1.如何查看模型的各个层次

print(model)#实例化后的

2.打印模型的参数

• 缺点：是可以看见一堆的参数，但是你不知道他是谁的参数，很乱。

print("-------------------------------------------------")
    print("model的bert的属性")
    for parameter in model.parameters():
        print(parameter)
        print("-------------------------------------------------")
    print("-------------------------------------------------")

3.打印模型的各层名字以及参数

print("-------------------------------------------------")
    print("model的bert的属性")
    for name,parameter in model.named_parameters():
        print(name,parameter)
        print("-------------------------------------------------")
    print("-------------------------------------------------")

     7.8309e-02, -2.3176e-02,  1.9839e-02, -1.7092e-02,  9.4321e-02,
    -1.4221e-02,  6.2530e-02, -3.1816e-02, -7.9080e-02,  1.4354e-02,
    -2.1350e-02, -5.6522e-02, -3.4564e-02], device='cuda:1',
   requires_grad=True)

classifier.weight Parameter containing:
tensor([[ 0.0106, -0.0667, -0.0263, …, -0.0182, -0.0581, 0.0074],
[ 0.0011, -0.0372, -0.0474, …, 0.0231, -0.0420, -0.0075],
[ 0.0103, 0.0349, 0.0343, …, -0.0065, -0.0091, 0.0024],
…,
[-0.1062, -0.1443, -0.0232, …, -0.0280, 0.0067, 0.0993],
[ 0.0186, -0.0386, 0.0207, …, -0.0693, -0.0363, 0.0977],
[ 0.0094, 0.0551, -0.0461, …, -0.0175, -0.0222, -0.0230]],
device=‘cuda:1’, requires_grad=True)

classifier.bias Parameter containing:
tensor([-0.3976, 0.1223, 0.3341, 0.3246, 0.2115, -0.0361, 0.2665, 0.1029,
0.4019, 0.1331, -0.1092, 0.1410, -1.5911, -0.1935, 0.2342],
device=‘cuda:1’, requires_grad=True)
4.返回模型的各种信息

for module in model.children():
     print(module)  # 打印网络第一代子模块
print("-------------------------------------------------")
for name, module in model.named_children():
     print(name, module)  # 打印模块名称 网络第一代子模块
print("-------------------------------------------------")
for module in model.modules():
     print(module)  # 打印模块
print("-------------------------------------------------")
for name, module in model.named_modules():
     print([name, module])  # 打印模块名称 模块
print("-------------------------------------------------")
print(model.state_dict())  # model.state_dict() 返回一个有序字典
print(model.state_dict().keys())
torch.save(model.state_dict(), "./weight.pth")
print("-------------------------------------------------")

weight = torch.load("./weight.pth")  # weight 是一个有序字典
print(weight)
# print("-------------------------------------------------")
for k, v in weight.items():
    print(k)

# model.modules()类似于 [[1, 2], 3],其遍历结果为：
[[1, 2], 3], [1, 2], 1, 2, 3
 
# model.children()类似于 [[1, 2], 3],其遍历结果为：
[1, 2], 3