torch.nn.functional.bilinear# torch.nn.functional.bilinear(input1, input2, weight, bias=None) → Tensor# 应用双线性变换到输入数据:y=x1TAx2+by = x_1^T A x_2 + by=x1TAx2+b 形状 input1:(N,∗,Hin1)(N, *, H_{in1})(N,∗,Hin1),其中Hin1=in1_featuresH_{in1}=\text{in1\_features}Hin1=in1_features,而∗*∗ 表示任意数量的额外维度。除了最后一个维度外,输入的维度应相同。 input2:(N,∗,Hin2)(N, *, H_{in2})(N,∗,Hin2),其中Hin2=in2_featuresH_{in2}=\text{in2\_features}Hin2=in2_features weight:(out_features,in1_features,in2_features)(\text{out\_features}, \text{in1\_features}, \text{in2\_features})(out_features,in1_features,in2_features) bias:(out_features)(\text{out\_features})(out_features) 输出: (N,∗,Hout)(N, *, H_{out})(N,∗,Hout) 其中 Hout=out_featuresH_{out}=\text{out\_features}Hout=out_features 并且除最后一个维度外,所有维度的形状都与输入相同。
torch.nn.functional.bilinear# torch.nn.functional.bilinear(input1, input2, weight, bias=None) → Tensor# 应用双线性变换到输入数据:y=x1TAx2+by = x_1^T A x_2 + by=x1TAx2+b 形状 input1:(N,∗,Hin1)(N, *, H_{in1})(N,∗,Hin1),其中Hin1=in1_featuresH_{in1}=\text{in1\_features}Hin1=in1_features,而∗*∗ 表示任意数量的额外维度。除了最后一个维度外,输入的维度应相同。 input2:(N,∗,Hin2)(N, *, H_{in2})(N,∗,Hin2),其中Hin2=in2_featuresH_{in2}=\text{in2\_features}Hin2=in2_features weight:(out_features,in1_features,in2_features)(\text{out\_features}, \text{in1\_features}, \text{in2\_features})(out_features,in1_features,in2_features) bias:(out_features)(\text{out\_features})(out_features) 输出: (N,∗,Hout)(N, *, H_{out})(N,∗,Hout) 其中 Hout=out_featuresH_{out}=\text{out\_features}Hout=out_features 并且除最后一个维度外,所有维度的形状都与输入相同。