blob: 782eec16a78be3078057f5175dca4c319f093b84 (
plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
|
#!/bin/bash -x
#####################################################################
# SUMMARY: Create a SentencePiece vocabulary with normalization rules
# AUTHOR: snukky
# TAGS: sentencepiece
#####################################################################
# Exit on error
set -e
# Remove old artifacts and create working directory
rm -rf vocab.norm vocab.norm.*{log,out,diff}
mkdir -p vocab.norm
# Run marian command
$MRT_MARIAN/marian \
--no-shuffle --seed 2222 --dim-emb 32 --dim-rnn 64 --maxi-batch 1 --maxi-batch-sort none --after-batches 1 \
-m vocab.norm/model.npz -t $MRT_DATA/europarl.de-en/corpus.small.{en,de}.gz \
--dim-vocabs 4000 -v vocab.norm/vocab.ende.spm vocab.norm/vocab.ende.spm --sentencepiece-options '--normalization_rule_tsv=norm.tsv' --sentencepiece-max-lines 10000 \
--log vocab.norm.log
# Check if files exist
test -e vocab.norm/model.npz
test -e vocab.norm/vocab.ende.spm
test -e vocab.norm.log
# Check logging messages
grep -q "Training SentencePiece vocabulary .*vocab.ende.spm" vocab.norm.log
# Extract a textual vocabulary and compare with the expected output
$MRT_MARIAN/spm_export_vocab --model vocab.norm/vocab.ende.spm > vocab.norm.out
$MRT_TOOLS/diff-nums.py vocab.norm.out vocab.norm.expected -o vocab.norm.diff
# Normalization is uppercasing, so check if there is no lowercased ASCII characters
grep -qvP '[a-z]' vocab.norm.out
# Exit with success code
exit 0
|
