Using REP legacy

2.1.0

Finish 2.1.0

Less dependencies, Python 2/3 compatible, new scores

Fully python 2 and 3 support

Less dependencies: install everything with [all]

train_similar_new contains better scores for comparison

Lighter interface: use rd.data.HEPDataStore, rd.stat., rd.ml. etc. to access the functions.

Stable for analysis

1.2.0

Finish 1.2.0