search

CambridgeCIA / LION

Learned Iterative Optimization Networks
GNU General Public License v3.0
20 stars 15 forks source link

Paths instruction and extension #96

Open mbkiss opened 6 months ago

mbkiss commented 6 months ago

I think it would be recommendable to write a short instruction on how to set up LION with changed paths:

Install conda environment

wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh -O conda.sh
bash conda.sh

Set up LION from own github fork on the gpu-server

git clone https://github.com/_CambridgeCIA_/LION.git (_or your own fork_)
cd LION
git submodule update --init --recursive
conda env create --file=env.yml
pip install git+https://github.com/ahendriksen/ts_algorithms.git **(because it fails before)**
conda activate LION

Change the paths in LION/LION/utils/paths.py and then move back to LION/ base pip install -e . Run the download.py file in LION/LION/data_loaders/2DeteCT/

  1. Maybe the paths in paths.py can be even set by default in reference to the installation folder
  2. In addition, I think it should include a path parameter for saving trained models and one for saving evaluations
  3. The last two mentioned paths could then be used in the standard templates for benchmarking and evaluation

This streamlines the individual usage of the toolbox where you only need to change a path in one location and not in many files.

AnderBiguri commented 6 months ago

100 agree!

I am hoping that we can have something like pip install -e , -path:"somepath", or something like this in the future, but this is good info in the meanwhile!

Your extra points are also noted, may be the way to go. Thanks!