2 releases
new 0.1.1 | Nov 14, 2024 |
---|---|
0.1.0 | Apr 1, 2024 |
#420 in Machine learning
121 downloads per month
Used in 2 crates
1MB
22K
SLoC
ZeNu Layer
ZeNu Layer is a collection of neural network layers implemented in Rust. It provides building blocks for constructing neural networks and integrates with the ZeNu deep learning library.
Features
- Various layer types, including fully connected (linear) layers
- Layer parameter initialization
- Forward pass computation
- Integration with ZeNu Autograd for automatic differentiation
Getting Started
To use ZeNu Layer in your Rust project, add the following to your Cargo.toml
file:
[dependencies]
zenu-layer = "0.1.0"
Here's a simple example of using a linear layer from ZeNu Layer:
use zenu_autograd::creator::from_vec::from_vec;
use zenu_layer::layers::linear::Linear;
use zenu_layer::Layer;
fn main() {
// Create a new linear layer with input dimension 3 and output dimension 2
let mut linear_layer = Linear::new(3, 2);
// Initialize the layer parameters with a random seed
linear_layer.init_parameters(Some(42));
// Create input data as a Variable
let input = from_vec(vec![1., 2., 3.], [1, 3]);
// Perform a forward pass through the layer
let output = linear_layer.call(input);
// Access the layer parameters
let parameters = linear_layer.parameters();
}
For more details and examples, please refer to the documentation.
License
ZeNu Layer is licensed under the MIT License.
Dependencies
~1.6–4MB
~73K SLoC