I'm working with MLFlow 2.7.1 on Windows 10 and when I try to serve a model via mlflow serve it tries to use Pyenv to install Python 3.10.12. I installed the latest version of Pyenv Win with pip and updated via pyenv update, but when I run pyenv install --list it does not list Python 3.10.12. Looking into it I found Pyenv Win does not seem to support Python 3.10.12 and the reason seems to be that python.org does not provide binaries/installers for Python 3.10.12 only source code. I tried modifying the mlflow\utils\virtualenv.py to use Python 3.10.11 by setting version='3.10.11' but then I got another error. Just wondering if anyone has any ideas on what I can do, or if there is some way I can specify to MLFlow to use a different version of Python than 3.10.12?
The command I am using to serve the model with MLFLow is;
mlflow models serve -m {path_to_model} -h 0.0.0.0 -p 8001
Try adding
--env-manager=localor--no-conda(deprecated) to your command:Refer to this post: Deploying MLflow Model without Conda environment