Uninstalling a preinstalled python package from Databricks

[KrishZ]

[Datasets](https://pypi.org/project/datasets/) python package comes preinstalled on databricks clusters. I want to uninstall it or completely prevent it's installation when I create/start a cluster.

I couldn't find any solution on stackoverflow.
And I have tried !pip uninstall datasets in a notebook. It throws the below error

```
Found existing installation: datasets 2.12.0
Not uninstalling datasets at /databricks/python3/lib/python3.10/site-packages, outside environment /local_disk0/.ephemeral_nfs/envs/pythonEnv-some-long-uu-id-I-have-hidden
Can't uninstall 'datasets'. No files were found to uninstall.
```

Can someone guide on what to do here ?

[Kaniz_Fatma]

Hi @KrishZ, To prevent Python packages from being installed when creating or starting a cluster, you can modify the cluster configuration by setting the spark.databricks.cluster.profile to serverless and spark.databricks.repl.allowedPyImportPrefixes To an empty string. This will turn off the installation of Python packages and prevent the execution of any Python code. Alternatively, you can use init scripts to remove any installed packages.

To uninstall a specific package, you can use the pip uninstall  command followed by the package name.

To uninstall the legacy Databricks CLI, you can use the command. pip uninstall databricks-cli.

[sean_owen]

@Kaniz_Fatma note that you can't actually uninstall packages in the runtime with pip.