Do we also plan on providing base fondant pytorch images for CUDA components? I guess there we'll have to make more limited choices (which pytorch version, python versions are limited)

I thought about this. I guess we don't need them right now. It can be quite complex and I think when people start using the GPU we should recommend building a custom component the old way. This components won't be lightweight neither.

Do we also plan on providing base fondant pytorch images for CUDA components? I guess there we'll have to make more limited choices (which pytorch version, python versions are limited)

I thought about this. I guess we don't need them right now. It can be quite complex and I think when people start using the GPU we should recommend building a custom component the old way. This components won't be lightweight neither.

alright makes sense! Some tests still to be failing though

Tested the build process in one of the latest commits.
Successfully executed run here: https://github.com/ml6team/fondant/actions/runs/7626293892/job/20772373591?pr=801

Btw I tested the base image in a local pipeline and it worked fine!

@lightweight_component(
    base_image="fndnt/fondant-base:dc8d9b208802ee85796238ff9a903c8bf66906e8-python3.8"
)
class FeatOne(PandasTransformComponent):
    def transform(self, dataframe: pd.DataFrame) -> pd.DataFrame:
        dataframe["one"] = dataframe["x"].map(lambda x: x + 1)
        return dataframe