1 unstable release
Uses new Rust 2024
| new 0.0.1 | Apr 18, 2025 |
|---|
#28 in #ode
75 downloads per month
29KB
601 lines
A minimalistic ODE solvers library built on top of PyTorch
Supported methods
| Solver | Method | Purpose / Use Case | Implicit | Adaptive step |
|---|---|---|---|---|
solve_euler |
Euler method | Simple and fast, good for teaching and very basic problems. | no |
no |
solve_rk4 |
Fourth Order Runge-Kutta (RK4) | Widely used general-purpose solver with fixed step size. | no |
no |
solve_implicit_euler |
Implicit Euler method | Suitable for stiff or ill-conditioned problems. | yes |
no |
solve_glrk4 |
Gauss-Legendre Runge-Kutta (Order 4) | High-precision solver for stiff or ill-conditioned problems. | yes |
no |
solve_rkf45 |
Runge-Kutta-Fehlberg 4(5) | Adaptive step size, balances efficiency and accuracy. | no |
yes |
solve_row1 |
First Order Rosenbrock-Wanner (ROW1) Method | Efficient alternative to Implicit Euler for stiff systems. | semi |
no |
Examples
Notebook example.ipynb has few examples of usage of mini-ode solvers.
Building Rust crate
In order to build Rust crate go to mini-ode/ directory and run command:
cargo build
Building Python package
Python package can only be built in conda environment or in python virtual environment. In order to build the project activate conda environment or python virtual environment. Then, go to mini-ode-python/ directory and run command:
LIBTORCH_USE_PYTORCH=1 maturin develop
Maturin will build the python package and install it in local environment.
Dependencies
~13MB
~263K SLoC