Unigram Agent

Baseline model which predicts the $n$ most common unigrams regardless of input.

Basic Examples

Training the unigram model on convai2:

parlai train_model -m unigram -mf unigram.model -t convai2 -eps 1 --num-words 15

DictionaryAgent Options

BPEHelper Arguments

Argument	Description
--bpe-vocab	Path to pre-trained tokenizer vocab
--bpe-merge	Path to pre-trained tokenizer merge
--bpe-dropout	Use BPE dropout during training.

UnigramAgent Options

optional arguments

Argument	Description
--num-words	Number of unigrams to output. Default: 10.

BPEHelper Arguments

Argument	Description
--bpe-vocab	Path to pre-trained tokenizer vocab
--bpe-merge	Path to pre-trained tokenizer merge
--bpe-dropout	Use BPE dropout during training.