OpenNMT Forum

Problem with update vocabulary in OpenNMT-py

Is there a way to update the vocabulary in OpenNMT-py. I’ve implemented and trained an NMT model and I want to fine-tune the model with another dataset.

To build the vocabulary, I’m using this command line:

!onmt_build_vocab -config config.yaml -n_sample -1

My config.yaml:

save_data: run/example
src_vocab: old_vocab.src
tgt_vocab: old_vocab.tgt
data:
corpus_1:
path_src: /content/new_en.txt
path_tgt: /content/new_es.txt
valid:
path_src: /content/devset.en
path_tgt: /content/devset.es

Discussed here: Finetuning with the new version OpenNMT-py v2.0 - #12 by francoishernandez