Expand description
An auto-difference library
Introduction
This is yet another auto-difference library for deep neural network. The focus is easy on use and dynamic computation graph building.
Install
Add auto-diff = “0.5” to the [dependencies] section of your project Cargo.toml file.
Features
The forward operators support a commonly used set, including:
- getter/setter,
- index and slicing,
- +, -, *, / and matmul,
- speciall functions,
- statistics,
- linear algebra,
- random number generator.
The corresponding gradient is work-in-progress.
One feature of auto-diff is the auto-difference is in background and don’t get in your way if only forward calculation is needed. Thus it can be used without syntax like variable place holder.
Example
Re-exports
pub use var::Var;pub use err::AutoDiffError;