kdmayer / PV_Pipeline

PyTorch models and pipeline developed for "DeepSolar for Germany". For reference, the paper can found at https://ieeexplore.ieee.org/document/9203258

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Reproducible environment with conda-lock

Baschdl opened this issue · comments

I cannot reproduce your environment on a Ubuntu machine using the requirements.txt

# platform: osx-64
affine=2.3.0=py_0
attrs=20.1.0=pyh9f0ad1d_0
backports=1.0=py_2
backports.functools_lru_cache=1.6.1=py_0

as it includes osx-specific build names for the packages. Could you generate a new environment.yml using conda-lock instead of conda-export as its use is discouraged for reproducible environments?

I tried to just remove the build numbers using https://stackoverflow.com/a/56780297/3342058 but that didn't work. I guess one would need a matching OS, python version and conda version to be able to reproduce it.

commented

Hi Baschdl,

Sure, I can look into that. In the meantime, I think it is easiest if you install the individual libraries manually. It should only be a handful of libraries that are actually needed to run this repo, i.e. torch, PIL, shapely, etc.

Hope that helps,
Kevin