update repo

嘉渊

2023-05-25 bade5bfca12fabd06271667c035c7fefc7b13982

 egs/wenetspeech/conformer/run.sh

@@ -20,7 +20,7 @@
type=sound
scp=wav.scp
speed_perturb="0.9 1.0 1.1"
stage=0
stage=1
stop_stage=5

# feature configuration
@@ -28,8 +28,7 @@
nj=64

# data
raw_data=../raw_data
data_url=www.openslr.org/resources/33
raw_data=/nfs/zhifu.gzf/wenetspeech_proc

# exp tag
tag="exp1"
@@ -42,7 +41,7 @@
set -u
set -o pipefail

train_set=train
train_set=train_l
valid_set=dev
test_sets="dev test_net test_meeting"

@@ -72,8 +71,15 @@
if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
    echo "stage 0: Data preparation"
    # Data preparation
    local/wenetspeech_data_prep.sh \
    --train-subset $set \
    $wenetspeech_data_dir \
    data || exit 1;
fi
    local/wenetspeech_data_prep.sh $raw_data $feats_dir
    mkdir $feats_dir/data
    mv $feats_dir/$train_set $feats_dir/data/$train_set
    for x in $test_sets; do
        mv $feats_dir/$x $feats_dir/data/
    done
fi

if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
    echo "stage 1: Feature and CMVN Generation"
    utils/compute_cmvn.sh --fbankdir ${feats_dir}/data/${train_set} --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} --config_file "$asr_config" --scale 0.1
fi