When we are using pytorch to build a model, we have to calculate the matrix product between 3D and 2D tensor. For example:
x = [batch_size, length, dim] = [32, 50, 200]
w = [dim, dim] = [200, 100]
y = f(x, w) = [32, 200, 100]
In this tutorial, we will introduce you how to do.
import torch x = torch.randn(32, 50, 200) w = torch.randn(200, 100) #[50,200] * [200,100] y = torch.matmul(x, w) print(y.shape)
Run this code, we can see:
torch.Size([32, 50, 100])
It means that we can use torch.matmul() function to calculate the the matrix product between 3D and 2D tensor in pytorch.
It is easy to do.