[Joss Review] Installation and Running Examples #2
Description
Hi @Mmasoud1, this issue is for openjournals/joss-reviews#9357. Here is some feedback to allow the installation process and running the example to be easier for the user:
- For the installation, I believe some required packages such as
nibabelare missing. This is imported in "meshnet_executor.py", so should be included during the installation proccess. - Maybe for the installation process, it would be best to list all the required Python packages in a requirements text file located in the root directory. This would allow a user to simply use `pip install -r requirements.txt" as opposed to installing each package separately in the wiki. The current "requirements.txt" file includes packages for testing so maybe this can be moved in the "/tests" folder?
- The installation instructions mentions checking the version for
nccbut doesn't mention how to install it. Maybe you can direct users to the instructions on how to install it via a link. For Linux users, I believe it's as simple assudo apt install nvidia-cuda-toolkit. The installation process for Windows users not using Windows Subsystem for Linux will be a different process. - For Python 3.12,
pip install nvflare==2.4.0, does not install, the documentation should note the versions of Python that work with the package. The environment example uses 3.9, which is compatible with the installation process. - The mindboggle databases in the examples directories point to the files stored in Git LFS. So users should be instructed to install Git LFS to get these files.
- For the example,
nvflare simulator -c site1,site2 ./jobs/job, should benvflare simulator -c site-1,site-2 ./jobs/job. - Also for the example, I believe there was an out of list error when collecting the gradients, I will have to run again and will update on the specific issue.
- Also, I believe the current "requirements.txt" file has an invalid
torchdistribution.
Thank you for your work on this package and I apologize for the late feedback!