You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm trying to set up the project on my system. However, the code does not run without additional changes. Furthermore, there are a lot of details missing from the readme.
Package installation initially failed at mpi4py-mpich due to a "No matching distribution found for mpi4py-mpich" error. I was able to resolve this by changing it to plain mpi4py in the requirements file. Also, at least on Windows, Python will still complain about missing libraries in mpi4py when the training script is run unless the user has installed MS-MPI prior to the mpi4py package. I have not checked to see if this affects other operating systems.
PyTorch and torchvision are missing from the requirements file.
I then got a "use_libuv was requested but PyTorch was build [sic] without libuv support" runtime error, and was only able to resolve it by downgrading PyTorch to version 2.3.0.
The readme does not mention that the user has to manually change path-to-glaucoma-detection-dataset to the data set folder in the shell scripts.
Python then threw an error saying that the "Distributed package doesn't have NCCL built in." I had to comment out the line dist_util.setup_dist() to proceed.
I also had to comment out the wandb initialization code due to an error about invalid API keys.
Python now throws a new RuntimeError because the path <data set>/train could not be found. The data set folder contains a lot of *.npz files, but it is not clear how to convert those to a train folder. Any advice would be appreciated.
The text was updated successfully, but these errors were encountered:
I'm trying to set up the project on my system. However, the code does not run without additional changes. Furthermore, there are a lot of details missing from the readme.
mpi4py-mpich
due to a "No matching distribution found for mpi4py-mpich" error. I was able to resolve this by changing it to plainmpi4py
in the requirements file. Also, at least on Windows, Python will still complain about missing libraries inmpi4py
when the training script is run unless the user has installed MS-MPI prior to thempi4py
package. I have not checked to see if this affects other operating systems.path-to-glaucoma-detection-dataset
to the data set folder in the shell scripts.dist_util.setup_dist()
to proceed.RuntimeError
because the path<data set>/train
could not be found. The data set folder contains a lot of *.npz files, but it is not clear how to convert those to atrain
folder. Any advice would be appreciated.The text was updated successfully, but these errors were encountered: