This is the code corresponding to the paper introducing EchoBERT a transformer-based model for interpreting echogram dynamics in farmed atlantic salmon.
The code provides two approaches for training networks using the Pytorch Lightning library, each in a separate main training code file:
- Train and test a single network (train_model_lightning.py)
- Train several networks at once using SLURM (distributed_job_manager.py)
It is possible to train both Transformer and LSTM networks by setting the ‘model_type’ argument in EchoBERT_Lightning.py. Similarly there are several other arguments that can be set in the two main training code files to load pretrained models or to decide whether to do pretraining or fine-tuning by setting the ‘disease_model’ argument to 0 or 1 respectively.
The dataset is available for download at: https://dx.doi.org/10.21227/76ma-tw16
The repo contains two .ipynb notebooks for creating the dataset from .csv files. First run Convert_to_.hdf5.ipynb to convert the .csv files to .hdf5 dataset and target files. Then run Make_x-fold.ipynb to create the different x-val folds from the dataset files.
The Convert_to_.hdf5.ipynb notebook must be run once per cage to convert the files, resulting in 6 runs per notebook (It’s not optimal I know). To create the six folds, each fold must be created manually by selecting which cage to exclude from the fold. This again results in a further six runs of the Make_x-fold.ipynb notebook. You should now be left with x-folds data and target files ready to be used for training.
The repo contains an anaconda environment.yml file which can be used to recreate the anaconda environment last used with this code.
Copyright (C) 2022 Håkon Måløy
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.