metadata

title: WildTorch
emoji: 🔥
colorFrom: red
colorTo: yellow
sdk: gradio
sdk_version: 4.26.0
app_file: app.py
pinned: false
license: mit

WildTorch

WildTorch: Leveraging GPU Acceleration for High-Fidelity, Stochastic Wildfire Simulations with PyTorch

GitHub: https://github.com/xiazeyu/WildTorch

Installation

Install with minimal dependencies:

pip install wildtorch

Install with full dependencies (includes visualization and logging):

pip install 'wildtorch[full]'

Quick Start

pip install 'wildtorch[full]'

import wildtorch as wt

wildfire_map = wt.dataset.generate_empty_dataset()

simulator = wt.WildTorchSimulator(
    wildfire_map=wildfire_map,
    simulator_constants=wt.SimulatorConstants(p_continue_burn=0.7),
    initial_ignition=wt.utils.create_ignition(shape=wildfire_map[0].shape),
)

logger = wt.logger.Logger()

for i in range(200):
    simulator.step()
    logger.log_stats(
        step=i,
        num_cells_on_fire=wt.metrics.cell_on_fire(simulator.fire_state).item(),
        num_cells_burned_out=wt.metrics.cell_burned_out(simulator.fire_state).item(),
    )
    logger.snapshot_simulation(simulator)

logger.save_logs()
logger.save_snapshots()

Demo

See Our Live Demo at Hugging Face Space.

API Documents

See at Our Read the Docs.