OFA-Visual_Grounding / fairseq /examples /multilingual /train_multilingual_model.sh
JustinLin610
update
10b0761
raw
history blame
1.3 kB
#!/bin/bash
# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
path_2_data=$1 # <path to data> which contains binarized data for each directions
lang_list=$2 # <path to a file which contains a list of languages separted by new lines>
lang_pairs=$3 #a list language pairs to train multilingual models, e.g. "en-fr,en-cs,fr-en,cs-en"
fairseq-train "$path_2_data" \
--encoder-normalize-before --decoder-normalize-before \
--arch transformer --layernorm-embedding \
--task translation_multi_simple_epoch \
--sampling-method "temperature" \
--sampling-temperature 1.5 \
--encoder-langtok "src" \
--decoder-langtok \
--lang-dict "$lang_list" \
--lang-pairs "$lang_pairs" \
--criterion label_smoothed_cross_entropy --label-smoothing 0.2 \
--optimizer adam --adam-eps 1e-06 --adam-betas '(0.9, 0.98)' \
--lr-scheduler inverse_sqrt --lr 3e-05 --warmup-updates 2500 --max-update 40000 \
--dropout 0.3 --attention-dropout 0.1 --weight-decay 0.0 \
--max-tokens 1024 --update-freq 2 \
--save-interval 1 --save-interval-updates 5000 --keep-interval-updates 10 --no-epoch-checkpoints \
--seed 222 --log-format simple --log-interval 2