Remove reliance on core_model_name

[sahiljhawar]

Remove core_model_name from model.py. Currently (if local flag is set) the model name and the SVD model names are supposed to be the same in order for it to work, however if one have multiple trained model files for same model, it is hard to use while keeping the models at one location. I want the model file to have any name and still work without relying on the modelname as the SVD directory name.

For example, if I want to use Bu2109lm model but I have two different version of the SVD i.e Bu2109lm_v1 and Bu2109lm_v2, I cannot use them unless I rename it to Bu2109lm.

Moving forward is this change worth it?

[mcoughlin]

@bking-astro Will likely have an opinion here given the LANL model naming struggles.

[bking-astro]

I think this should be fine to change, as long as it's not a default setting to specify different names for model and svd dir. However I have be just making separate directories denoting the version or type of training with the typical "BU2019lm" model dir inside of it.

[sahiljhawar]

making separate directories denoting the version or type of training with the typical "BU2019lm" model dir inside of it.

@bking-astro Can you explain more?

[bking-astro]

I would just have, for example, /lustre/scratch5/bking/svdmodels/Bu2019lm_smoothed_15/Bu2019lm/Bu2019lm_tf.pkl and my svd_path would be /lustre/scratch5/bking/svdmodels/Bu2019lm_smoothed_15/.