🤕Troubleshooting
Last updated
Last updated
Open the app's settings page and check if the Go server is running:
If the server is not running click "Start server" to restart it:
If none of the aforementioned steps help solve your problem, please contact us.
Your python environment could have problems.
You can check your python environment's path using the following:
For windows: use the command where python
and select the path in the .medomics
folder.
For Linux and MACOS: use the command whereis python
and select the one inside .medomics
folder. If you don't see any path with the .medomics
folder, find it manually and export it to your PATH
environment variable using the command: export PATH=/path/to/your/usr/.medomics/python/bin:$PATH
If none of the aforementioned steps help solve your problem, please contact us.
Problem description:
The app seems to be stuck at startup. Or after startup, when opening multiple tabs (e.g. two different csv files), the app turns all white.
Probable cause:
If you have an insitutitonal antivirus like Trellix at Université de Sherbrooke, we believe the firewall of Trellix prevents MEDomicsLab to run properly. We hope the problem will be solved once we certify (signing Mac software) the MEDomicsLab app with Apple (work in progress).
Workaround:
If you encounter a similar issue, and you have an antivirus installed on your computer, we recommend that you disable it when using MEDomicsLab. For Mac users at Université de Sherbrooke with Trellix, one way to momentarily disable the utilities of Trellix is to:
Go to System Settings, Privacy & Security, Full Disk Access.
Disable the following six extensions: fmpd, VShieldScanManager, VShieldScanner, masvc, TrellixEndpointSecurity, TrellixNetworkExtension.
Restart yur computer.
Of course, once you are done working with MEDomicsLab, we recommend you enable again your antivirus! For Mac users at Unversité de Sherbrooke with Trellix, this would mean to enable again the six extensions above, and to restart your computer.
This error occurs when the wrong version of dtreeviz is installed. Please refer to the requirement files in our GitHub repository to install the right version (v1.4.1 is the one usually used).
The error message typically appears as ['column_1', 'column_2', 'column_3'] not in index
. This issue occurs when column names in the file used for model training contain spaces (e.g., "column 1" instead of "column_1"). To prevent this, avoid using column names with spaces. While the code should handle this automatically, please open an issue on our GitHub if you encounter this error.
This error usually occurs due to a missing Python library (shown as a ModuleNotFoundError
). To resolve it, first open the console (press CTRL + Shift + I) and identify the missing package name in the full error message. Then, install the specified package version from the requirements file on our GitHub repository. For a more comprehensive solution, compare your installed packages with those listed in the requirements file on GitHub and install any missing packages to prevent this error in the future.
If the missing package is not listed in our GitHub requirements, please contact us or open an issue so we can add it.
This is a common issue in Ubuntu that occurs when training a lightGBM model using n_jobs=-1
. When you set n_jobs=-1
, the process can sometimes hang or slow down due to how the library manages parallel threads on certain systems. Here are some common reasons for this behavior:
CPU Thread Saturation: Setting n_jobs=-1
tells LightGBM to use all available CPU cores. On Linux systems, particularly with high core counts, this can overwhelm the CPU scheduler, leading to inefficient thread management, especially if other processes are running concurrently.
OpenMP Configuration: LightGBM uses OpenMP for parallelism, and certain configurations of OpenMP on Linux can lead to deadlocks or excessive thread contention. This issue can be specific to how OpenMP handles threads in certain Linux distributions, including Ubuntu. More details here: https://lightgbm.readthedocs.io/en/latest/FAQ.html#lightgbm-hangs-when-multithreading-openmp-and-using-forking-in-linux-at-the-same-time
Memory Bandwidth and Latency: Using all CPU cores can lead to high memory bandwidth consumption. If LightGBM needs more memory per thread than available, this can slow down training significantly. Lowering the n_jobs
setting reduces the number of simultaneous threads and can help manage memory load.
Compatibility with Specific Libraries: LightGBM's multithreading may not work smoothly with all versions of system libraries on Ubuntu, such as libgomp
(GNU OpenMP). Sometimes, upgrading or downgrading certain libraries (e.g., libgomp1
) can resolve this issue.
Limit n_jobs
: Set n_jobs
to a smaller number (e.g., 4 or 8), which often yields good performance without overloading the system.
Upgrade/Downgrade System Libraries: Update your OpenMP libraries.
You can test different n_jobs
values to find the optimal setting, which balances speed and stability.
Occurs when using XGBoost or catboost.
The error occurs because the library is not installed or configured in your environment.
Solution: Install the missing library using pip, for example:
pip install catboost
However, you must install the missing library in the specific MEDomicsLab bundled Python in your user folder, replace YourUsername
with the full path to your Python executable when running the pip command. Here's an example:
Locate your bundled Python executable (e.g., C:\Users\<YourUsername>\.medomics\python\python.exe
).
Run the following command (assuming catboost is the missing package ):
C:\Users\<YourUsername>\.medomics\python\python.exe -m pip install catboost
This ensures the library is installed in the MEDomicsLab Python environment.
This error is encountered when trying to open a Jupyter Notebook using VSCode:
The cause is simply the "code" command missing. Follow the instructions here to add it.
In a terminal, write the following command:
sudo apt remove medomicslab-application
Go in your Applications Folder in Finder.
Then, click on the MEDomicsLab Icon while holding the Ctrl
key.
Finally, click on "Move to Trash"
Go to Settings > Apps
Then, click on "Installed Apps"
Search for "MEDomicsLab"
Click on the ...
and finally click on "Uninstall"
On any operating system, to remove the MEDomicsLab Python environment, go to your user's folder (for example C:\Users\alex
) and manually remove the .medomics
folder