I am trying to deploy either FB Prophet or FB NeuralProphet to the Streamlit Cloud. I have been trying with NeuralProphet because it appears to have fewer problematic dependencies. I tried both just specifying the package in the requirements.txt file:
neuralprophet==0.4.1
That blew up. So then I tried specifying all the dependencies in the order they install when you pip install neuralprophet. That also blew up. Any ideas??
I was able to deploy NeuralProphet. I imported it without a version number and that seemed to work.
I still can’t get fbprophet to load, which is frustrating. But I always have issues installing the pystan packages locally.
If anyone knows any easier packages to use on Streamlit Cloud for forecasting sales from a timeseries, I am all ears! I just like fbprophet’s automatic hyperparameter adjustments.
Hi, you should try to pip install prophet==1.1.1
(and not fbprophet) to avoid pystan issues.
I have deployed it on this app:
https://prophet.streamlit.app/
1 Like