library_structure.md

← Library Structure

The library is structured as follows:

├── README.md
│
├── config
│    │
│    ├── experiments
│    │   ├── UPP_16.toml
│    │   ├── UPP_32.toml
│    │   ├── UPP_64.toml
│    │   ├── UPP_all.toml
│    │
│    ├── configuration.toml
│    ├── discord.toml
│    ├── models.toml
│    ├── tflow.toml
│    ├── torch.toml
│    ├── train.toml
│
├── data
│   ├── sfv03_fcci.zarr
│   │
│   ├── scaler
│   │   ├── fcci_max_point_map.nc
│   │   ├── fcci_min_point_map.nc
│   │   ├── fcci_mean_point_map.nc
│   │   ├── fcci_stdv_point_map.nc
│
├── digital_twin_notebooks
│   │
│   ├── img
│   ├── inference_on_test_data.ipynb
│   ├── inference_with_onnx.ipynb
│   ├── model_conversion_pytorch_to_onnx.ipynb
│
├── docs
│
├── experiments
│   │
│   ├── 20240311_222733
│   │   ├── exp_5.toml
│   │   ├── fabric_benchmark.csv
│   │   ├── last_model.onnx
│   │   ├── last_model.pt
│
├── Fires
│   │
│   ├── _datasets
│   │   ├── dataset_zarr.py
│   │   ├── torch_dataset.py
│   │
│   ├── _layers
│   │   ├── unetpp.py
│   │
│   ├── _macros
│   │   ├── macros.py
│   │
│   ├── _models
│   │   ├── base.py
│   │   ├── unetpp.py
│   │   ├── vgg.py
│   │
│   ├── _scalers
│   │   ├── base.py
│   │   ├── minmax.py
│   │   ├── scaling_maps.py
│   │   ├── standard.py
│   │
│   ├── _utilities
│   │   ├── callbacks.py
│   │   ├── cli_args_checker.py
│   │   ├── cli_args_parser.py
│   │   ├── configuration.py
│   │   ├── decorators.py
│   │   ├── logger.py
│   │   ├── swin_model.py
│   │
│   ├── __init__.py
│   ├── augmentation.py
│   ├── datasets.py
│   ├── layers.py
│   ├── macros.py
│   ├── models.py
│   ├── scalers.py
│   ├── trainer.py
│   ├── utils.py
│
├── main.py
├── launch.sh

File	Type	Main function
config	$\small{\textcolor{blue}{\texttt{DIR}}}$	It stores TOML configuration files
data	$\small{\textcolor{blue}{\texttt{DIR}}}$	It stores important files (e.g., scalers, symbolic links to the dataset, etc)
digital_twin_notebooks	$\small{\textcolor{blue}{\texttt{DIR}}}$	It stores Jupyter notebooks that carry on the Digital Twin's tasks on Wildfires use case
docs	$\small{\textcolor{blue}{\texttt{DIR}}}$	It stores documentation files
experiments	$\small{\textcolor{blue}{\texttt{DIR}}}$	It contains folders named with the current date and time when the experiment took place. The best model, the experiment configuration file and the benchmark file will be saved in this folder after the completion of the experiment.
Fires	$\small{\textcolor{blue}{\texttt{DIR}}}$	It is the main library that is used to carry on the training of the Machine Learning model and the inference on the SeasFireCube data. It is used to store, in an organized way, all the code that provides support to the main.py script during its execution. Model implementations and training utility functions can be found here. (details)
main.py	$\small{\textcolor{red}{\texttt{FILE}}}$	It contains all the workflow code that must be executed.
launch.sh	$\small{\textcolor{red}{\texttt{FILE}}}$	It runs the experiments once it has been executed (details)

Warning

Both data and experiments directories are empty and are not included in the repository as they contain too heavy files that cannot be stored.

Tip

Before running the code, remeber to download the SeasFire Cube v3 and put into data folder.