Burn Core: burn::tensor::Tensor::matmul
The Tensor::matmul method performs matrix multiplication between two compatible tensors. The left-hand tensor supplies the row dimension, and the right-hand tensor supplies the output column dimension. For rank-2 tensors, the inner dimensions must match.
The returned tensor has the resulting matrix shape [m, n] for standard two-dimensional multiplication. In model code, this operation is commonly used inside dense layers, attention projections, and linear transformations.
Parameters
| Parameter Name | Type | Description |
|---|---|---|
self |
Tensor<B, 2> |
The left-hand matrix tensor with shape [m, k]. |
other |
Tensor<B, 2> |
The right-hand matrix tensor with shape [k, n]. |
let device = B::Device::default();
let a = Tensor::<B, 2>::ones([2, 3], &device);
let b = Tensor::<B, 2>::ones([3, 4], &device);
let y = a.matmul(b);Example Output
{
"status": "success",
"operation": "matmul",
"input_shape": [2, 3],
"rhs_shape": [3, 4],
"output_shape": [2, 4]
}