search

rdadolf / fathom

Reference workloads for modern deep learning methods.
Apache License 2.0
73 stars 21 forks source link

Flat renaming #3

Closed rdadolf closed 8 years ago

rdadolf commented 8 years ago

I'd like some structural cleanup. Specifically:

  1. Flat names. I'd like each model to have its own top-level directory. This is to make life easier for users.
  2. Export models to the top level.
  3. Standardization for dataset selection and training/inference modes. We need the ability to name a model which uses a small, debug dataset (#4) or the large, real dataset. We also need to be able to name the training and inference models (which we currently can).
rdadolf commented 8 years ago

8d103e9 fixes 1. and 2.

7 is a better solution to 3.