Hi All,

I wanted to share an issue with you guys regarding the ML app we had in our server in which the app was automatically changing it's python package versions despite of whatever desired versions of python packages we install. Then we figured out that if the ML app had a file named virtual environments and in it's metadata if the versions are already set , no matter what ever desired versions of python packages we install it automatically changes it's version to the one mentioned in the metadata of that virtual environment. The reason for this is that the workbenches where virtual environments were used usually become corrupted and inaccessible. Files will not be recoverable if a workbench is corrupted. 

The only way of resolving this issue is creating a new ML app and backing up the files in the current workbench (Excluding the virtual environments) and restoring them to a new workbench. This will ensure your data is protected from possible corruption. For more information regarding moving your files to workbench please refer the article below.

Machine Learning: Backup a Workbench (MLWB) Manually (celopeers.com)