math nlp

Arithmetic

This is a really simple little encoding with d_vocab = 14 (0-9, +, -, =, >), where '>' is the BOS/EOS token.

Models trained with this should be incoming soon-ish.