Getting Started

Prerequisites: Python 3.12+. PyTorch is optional, required only for examples that import torch.nn.

Install

pip install laco
# Optional: with wandb integration
pip install "laco[wandb]"

Five-Minute Example

1. Write a config file

# configs/mlp.py
import laco.language as L
import torch.nn as nn

model = L.call(nn.Sequential)(
    L.call(nn.Linear)(in_features=784, out_features=256),
    L.call(nn.ReLU)(),
    L.call(nn.Linear)(in_features=256, out_features=10),
)

2. Load and inspect

import laco

cfg = laco.load("configs/mlp.py")
print(laco.dump(cfg))  # prints Hydra-compatible YAML

3. Instantiate

model = laco.instantiate(cfg.model)
# model is now a real nn.Sequential
print(type(model))  # <class 'torch.nn.modules.container.Sequential'>

4. Apply overrides

cfg = laco.load("configs/mlp.py?model.0.out_features=512")

Or from the CLI:

laco compose configs/mlp.py model.0.out_features=512

Next Steps

Work through the tutorials: notebooks/tutorials/
Understand the lie-typing contract
See full examples in examples/foundations.md
Browse the API reference

Learn