Burn Core: burn::tensor::Tensor::unsqueeze
The Tensor::unsqueeze method increases the rank of a tensor by adding leading dimensions. The type parameter defines the resulting tensor rank. This operation is useful when a lower-rank tensor must be broadcast against a higher-rank tensor.
Unsqueezing does not alter the underlying scalar values. It changes the dimensional interpretation so that shape-compatible operations can proceed. The resulting rank must be greater than or equal to the original rank.
Parameters
| Parameter Name | Type | Description |
|---|---|---|
self |
Tensor<B, D> |
The source tensor whose rank will be expanded. |
D2 |
const generic usize |
The resulting tensor rank after inserting additional dimensions. |
let device = B::Device::default();
let bias = Tensor::<B, 1>::ones([8], &device);
let expanded = bias.unsqueeze::<2>();
println!("{expanded}");Example Output
{
"status": "success",
"operation": "unsqueeze",
"input_rank": 1,
"output_rank": 2
}