Skip to content

Latest commit

 

History

History
86 lines (70 loc) · 2.72 KB

README.md

File metadata and controls

86 lines (70 loc) · 2.72 KB
Raw

AutoPhaseNN

Link to the application: https://github.com/YudongYao/AutoPhaseNN

References:

How to use (training)

Setup and download dataset

  1. Get the SBI-FAIR repository

    git clone --depth 1 https://github.com/DSC-SPIDAL/sbi-fair
SBI_FAIR_DIR=${PWD}/sbi-fair

  2. Create a directory for downloading datasets and store results

    mkdir autophasenn
cd autophasenn
mkdir output

  3. Get the datasets for training

    The AutoPhaseNN dataset (a list of files) requires processing (downloading data and upscaling). The loading script will build and use the Docker container that is provided in the repository to do so.

    Instead of the whole dataset you may want to load some sample files for testing. If so, before running the loading script create a shorter list of files for downloading:

    mkdir -p aicdi/default
wget -q -O - https://github.com/YudongYao/AutoPhaseNN/raw/3375cf98206a83f329faaf4c74eed924f3f4a2fe/TF2/aicdi_data.txt | head -n 300 > aicdi/default/aicdi_data.txt
    ${SBI_FAIR_DIR}/tools/scripts/load_dataset.py ${SBI_FAIR_DIR}/datasets/autophasenn/datasets.yaml aicdi

  4. Create a file with parameters

    # Few epochs for testing
echo 'epochs: 2' > options.yaml 
echo 'gpu_num: 1' >> options.yaml 
echo 'train_size: 100' >> options.yaml

    We will update the list of available options here, in the meantime please refer to the original repository https://github.com/YudongYao/AutoPhaseNN for the list of all options.

Using Docker

  1. Build Docker container

    cd ${SBI_FAIR_DIR}/models/autophasenn
./build_docker.sh
cd - # Go back to results directory

  2. Run Training

    GPU_SWITCH='--runtime=nvidia --gpus all' # or '' for CPU workloads
# Mount the directories with the dataset
VOLUME_MOUNTS='-v ./aicdi/default:/input/train_dataset -v ./output:/output -v ./options.yaml:/input/options.yaml'
docker run ${GPU_SWITCH} ${VOLUME_MOUNTS} autophasenn run train

Using Apptainer

  1. Build Apptainer container

    cd ${SBI_FAIR_DIR}/models/autophasenn
./build_apptainer.sh
cd - # Go back to results directory

  2. Run Training

    GPU_SWITCH='--nv' # or '' for CPU workloads
# Mount the directories with the dataset
VOLUME_MOUNTS='--bind ./aicdi/default:/input/train_dataset --bind ./output:/output --bind ./options.yaml:/input/options.yaml'
apptainer run --app train ${GPU_SWITCH} ${VOLUME_MOUNTS} ${SBI_FAIR_DIR}/autophasenn/autophasenn.sif

Results

The outputs of the run will be available in ./output.