search

jrzaurin / pytorch-widedeep

A flexible package for multimodal-deep-learning to combine tabular data with text and images using Wide and Deep models in Pytorch
Apache License 2.0
1.3k stars 190 forks source link

Consistent names for model components #34

Closed jrzaurin closed 3 years ago

jrzaurin commented 3 years ago

At the moment there are two inconsistencies of different nature:

  1. There is one very small inconsistency between pypi and the github version. If you installed the pypi version, the Transformer Encoder blocks component of the TabTransformer model are called blks. However, if you install from github those blocks are called tab_transformer_blks, which is a name more consistent with the other two models (TabMlp and TabResnet). This inconsistency will be fixed in the coming release
  2. The Resnet blocks in the TabResnet model are named tab_resnet. A more consistent name would be tab_resnet_blks

All this needs to be fixed in the next version

jrzaurin commented 3 years ago

Done