search

Fernando-Urbano / monash-time-series-replication

Monash Time-Series Forecast in Reproducible Analytical Pipeline (RAP) format
0 stars 3 forks source link

1.1. Create virtual env #8

Closed Fernando-Urbano closed 6 months ago

Fernando-Urbano commented 6 months ago

Create virtual env that has Python and R

Fernando-Urbano commented 6 months ago

It was way harder than I thought to generate a proper virtual environment.

First, the R and Python that were used in their code was not available in conda. That generates further problems to install the packages that they used in the right version.

Furthermore, they use 4 special packages: gluonts, forecast, glmnet, smooth.

Because of the problems in virtual environments, I had to download versions that are not the same as the ones used in project. Therefore, if you at anytime see that the result is diverging from the original paper, please text me.

I tried multiple solutions yesterday, but none was perfect in terms of versioning. I went with the most recent versions of R and Python because they are generating way less trouble when trying to download packages

I will continue to try to create a virtual env that is equal to theirs, but this MVP should already work.

Fernando-Urbano commented 6 months ago

Aben did a lot of that: it took a considerable time to find the environment that would work for the R and python models and assure that everything was good. Cheers to aben!