Metadata-Version: 2.4
Name: rippl
Version: 0.6.7
Summary: High-performance physics compiler and PINN engine.
Requires-Python: >=3.10
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: torch>=2.0
Requires-Dist: pydantic
Provides-Extra: dynamics
Requires-Dist: torchdiffeq>=0.2.3; extra == "dynamics"
Provides-Extra: test
Requires-Dist: pytest; extra == "test"
Requires-Dist: safetensors; extra == "test"
Requires-Dist: scipy; extra == "test"
Dynamic: license-file

# rippl
**The physics compiler for neural networks.**

[![PyPI](https://img.shields.io/pypi/v/rippl)](https://pypi.org/project/rippl)
[![License](https://img.shields.io/badge/license-Apache%202.0-blue)](LICENSE)

→ **[Full Documentation](docs/)** · [Benchmarks](rippl/benchmarks/README.md)

---
## What it is
rippl is a physics compiler. You write the PDE. rippl solves it.
It sits between you and PyTorch — handling residual computation, 
boundary enforcement, causal training, and adaptive loss balancing 
so you don't have to.

---
## 7 Lines
```python
import rippl
import rippl.nn as nn

model  = nn.MLP(in_dim=2, out_dim=1)
engine = rippl.compile(model)
result = rippl.run(domain, equation, engine,
                causal=True, adaptive_loss=True, hard_bcs=True)
```

---
## Install
```bash
pip install rippl
```

---
## Multi-GPU Scaling
Multi-GPU DDP training and advanced 3D mesh ingestion are supported via the `rippl-pro` plugin architecture. 

---
## License
Apache 2.0 — see [LICENSE](LICENSE)
