LanguageDetection/onmt_train.sh at master · pablogadhi/LanguageDetection · GitHub

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
data_dir="./data/*"

export CUDA_VISIBLE_DEVICES=0

steps=$3
val_steps=$4
batch_size=$5
save_at=$6
checkpoint=$7

if [ -z "$steps" ]; then
    steps=200000
fi

if [ -z "$val_steps" ]; then
    steps=10000
fi

if [ -z "$batch_size" ]; then
    batch_size=4096
fi

if [ -z "$save_at" ]; then
    save_at=10000
fi

onmt_train -data $1 -save_model $2 \
    -layers 6 -rnn_size 512 -word_vec_size 512 -transformer_ff 2048 -heads 8 \
    -encoder_type transformer -decoder_type transformer -position_encoding \
    -train_steps $steps -max_generator_batches 2 -dropout 0.1 \
    -batch_size $batch_size -batch_type tokens -normalization tokens -accum_count 2 \
    -optim adam -adam_beta2 0.998 -decay_method noam -warmup_steps 8000 -learning_rate 2 \
    -max_grad_norm 0 -param_init 0 -param_init_glorot \
    -label_smoothing 0.1 -valid_steps $val_steps -save_checkpoint_steps $save_at \
    -world_size 1 -gpu_ranks 0 -train_from "$checkpoint" -keep_checkpoint 20