DIVERS-H Protos

Started Feb 2025

neuralhydrology

Github workflow has neuralhydrology-image.yml to build docker image of the Google NeuralHydrology package.

To build locally:

cd neuralhydrology VERSION=$(cat neuralhydrology.version.txt) && docker build -f neuralhydrology.dockerfile -t neuralhydrology/base:${VERSION} .

neuralhydrology/demo1

To build locally:

cd neuralhydrology/demo1 VERSION=$(cat demo1.version.txt) && docker build -f demo1.dockerfile -t neuralhydrology/demo1:${VERSION} .

Local demo for single basin train & test

cd neuralhydrology python demo1/local_main.py -s train -b 02430085 python demo1/local_main.py -s test -b 02430085 -r run_1502_194916

Will read .args.txt file if present. File contents, ,e.g.:

--data_dir
/home/john/projects/divers-h/data/camels/basin_dataset_public_v1p2
--experiments_root_dir
/home/john/projects/divers-h/experiments

Containerized neuralhydrology/demo1

To build cd neuralhydrology/demo1 docker build -f demo1.dockerfile -t nh/demo1 .

To start container:

docker run --rm -it --gpus all \
  --mount type=bind,src=/home/john/projects/divers-h/data/camels/basin_dataset_public_v1p2,dst=/data,readonly \
  nh/demo1 \
  bash

Then:

• python local_main.py -s train -d /data -e /experiments -b 02430085
• python local_main.py -s test -d /data -e /experiments -b 02430085 -r <RUN_ID>

To run the standalone demo: cd neuralhydrology python demo.py -s train -d /data -e /experiments -b 02430085 -r <RUN_ID>