Closed thistlillo closed 1 day ago
Hey @thistlillo Franco 👋
Thank you for bringing this to our attention!
I have created a ticket for the dev team to update the conda recipe for neptune-optuna
, and I will update this thread once it is done ✅
Meanwhile, is it possible for you to use pip
to install packages?
If yes, you can uninstall the conda installations of neptune
, neptune-client
, and neptune-optuna
by running the below:
conda uninstall neptune neptune-client neptune-optuna -y
and then just install neptune
and neptune-optuna
using pip
as below:
pip install -U neptune[optuna]
Please let me know if this works for you.
We are not distributing on conda anymore.
Would request you to install via pip.
Ok, thank you. That is what I do, but it leads to messy situation.
Please notice that on remote servers your users do not usually have sudo access and it is almost impossible to work without a simple way to change, for example, the current python version available on those systems. In these cases, pip does not help.
Pip should take the python version into account when installing packages. Is that not the case with your setup? 🤔
No, I am saying that sometimes I need a specific Python version for my ML pipeline, which I can install through conda. That's what I have been doing so far. Installing/uninstalling libraries in the same conda env using both conda and pip has led (neither so often nor once or twice) to a complete mess.
Describe the bug
I would like to log on neptune the results of a hyperparameter sweep using optuna. I follow the instructions on: https://docs.neptune.ai/integrations/optuna/
I tried several installation sequences, they all lead to failure.
Reproduction
conda install neptune neptune-optuna then follow the examples on the Neptune website
Expected behavior
Not such an error.
Traceback
Environment
This is the output of
conda env export
linux
Python 3.9.16