CIP_-MoDL_vs_VarNet-

Computational Imaging Project CIP

References:

• MoDL source code: https://github.com/hkaggarwal/modl

• VarNet source code: https://github.com/facebookresearch/fastMRI

"Blind" tutorial using HPC

1. login

• ssh <user>@tinyx.nhr.fau.de

2. Write in the terminal

• module load python/3.8-anaconda

• conda init bash

• source ~/.bashrc

• nano ~/.profile

•   if [ -n "$BASH_VERSION" ]; then
    #  include .bashrc if it exists
    if [ -f "$HOME/.bashrc" ]; then
    . "$HOME/.bashrc"
    fi
    fi 
  
  

• conda config # create ~/.condarc

• nano ~/.condarc

• pkgs_dirs:
  - ${WORK}/software/privat/conda/pkgs
  envs_dirs:
  - ${WORK}/software/privat/conda/envs
  
  
  

• conda create --name myenv

• conda activate myenv

• pip install tensorflow matplotlib numpy fastmri h5py tqdm

• NOTE: If some of the packages can't be pip instal try using conda install <package>

3. USE your HPC account using TinyGPU on https://jupyterhub.rrze.uni-erlangen.de/

4. Select the kernel named Python [conda env:conda-myenv]

5. Link the folder to training data (you have to be in CIP_-MoDL_vs_VarNet/):

• ln -s /home/vault/iwbi/shared/cip22_varnet_modl/brain/train
• ln -s /home/vault/iwbi/shared/cip22_varnet_modl/brain/test
• ln -s /home/vault/iwbi/shared/cip22_varnet_modl/brain/val

6. Usage Compat:

• tf_upgrade_v2 --infile <module_to_convert> --outfile <name_of_the_target_file>

• See Upgrading your code to Tensorflow 2.0

7. Copying files from HPC to local

• scp -r <user>@tinyx.nhr.fau.de:/home/hpc/iwbi/<user>/origin_dir home/user/objective_directory