yufengwa / fine-tuning-nos

Fine-tuning Neural-Operator architectures for training and generalization.

This is the official implementation of the paper "Fine-tuning Neural-Operator architectures for training and generalization".

Get started (Libraries)

If you want to reproduce all the results (including the baselines) shown in the paper,

You can then set up a conda environment with all dependencies like so:

conda env create -f environment.yaml
conda activate forward-operator

Dataset

The data set is proportioned upon request. It must be located in a directory with the following structure:

databases/acoustic/GRF_{Freq}Hz/data
databases/acoustic/GRF_{Freq}Hz/model

In the place holder is $7, 12, 15$ Hz the Frequencies of the experiments. The configuration for the experiments are located in

dataset_time-harmonic-waves_hawen_parameters

Configuration of architectures

All the architectures' parameters are located in /config directory.

config/acoustic/GRF_{Freq}Hz/<Architecture>.yaml

Traning

They can be trained with

CUDA_VISIBLE_DEVICES={k} python3 main.py -c <path_to_config_file>

Evaluation

We train multiple times in the code. The evaluation is a function of the amoung of saving files. It can be implemented as follows

CUDA_VISIBLE_DEVICES={k} python3 evaluation.py -n <number_training_save_model> -c <path_to_config_file>

Plot (no visualization)

The same structure follows for plotting

 python3 reconstruction_plot.py -c <path_to_config_file>

Loss Landscape visualization

visualization_code