PyTorch的矩阵乘法与线性变换

        #torch.nn下的Linear是线性变换，包括矩阵乘法和偏置量

        import torch
        import torch.nn as nn


        batch_size = 4
        input_features = 4
        output_features =3

        x=torch.randn(batch_size, input_features)

        linear_layer = nn.Linear(input_features, output_features)   #自动生成随机的权重矩阵weight,和偏置bias
        out_linear = linear_layer(x)


    
        

    #使用matmul
        weight = linear_layer.weight
        bias = linear_layer.bias
        out_matmul = x @ weight.T +bias

        print(out_linear)
        print(out_matmul)
        print("Is equal=", torch.allclose(out_linear, out_matmul))

        # tensor([[ 0.1254,  0.6651, -0.3266],
        #         [ 0.0621,  0.8397,  0.6082],
        #         [-0.5628,  0.0811,  0.4234],
        #         [ 0.6737,  0.0293, -0.2010]], grad_fn=)
        # tensor([[ 0.1254,  0.6651, -0.3266],
        #         [ 0.0621,  0.8397,  0.6082],
        #         [-0.5628,  0.0811,  0.4234],
        #         [ 0.6737,  0.0293, -0.2010]], grad_fn=)
        # Is equal= True