zpwiki/pages/students/2016/patrik_pavlisin/dp2022/train1.sh

33 lines
921 B
Bash
Raw Normal View History

# train.yaml
## Where the samples will be written
save_data: dp2022/run2/example
## Where the vocab(s) will be written
src_vocab: dp2022/run2/example.vocab.src
tgt_vocab: dp2022/run2/example.vocab.tgt
# Prevent overwriting existing files in the folder
overwrite: False
# Corpus opts
data:
corpus_1:
path_src: dp2022/europarl-v7.sk-en.en
path_tgt: dp2022/europarl-v7.sk-en.sk
transforms: [onmt_tokenize]
weight: 1
valid:
path_src: dp2022/europarl-v7.clean.sk-en.en
path_tgt: dp2022/europarl-v7.clean.sk-en.sk
transforms: [onmt_tokenize]
# Vocabulary files that were just created
src_vocab: dp2022/run2/example.vocab.src
tgt_vocab: dp2022/run2/example.vocab.tgt
# Where to save the checkpoints
save_model: dp2022/run2/model
save_checkpoint_steps: 1000
train_steps: 20000
valid_steps: 10000