update repo

嘉渊

2023-05-11 23a749f956affdfe078f8b8eeaa59b830efb5ff7

update repo

 egs/aishell/data2vec_paraformer_finetune/conf/train_asr_paraformer_transformer_12e_6d_3072_768.yaml

@@ -30,7 +30,6 @@
    require_same_masks: true
    mask_dropout: 0


# decoder related
decoder: paraformer_decoder_san
decoder_conf:
@@ -53,6 +52,7 @@
    lfr_m: 1
    lfr_n: 1

# hybrid CTC/attention
model: paraformer
model_conf:
    ctc_weight: 0.3

 egs/aishell/data2vec_paraformer_finetune/local/download_and_untar.sh

New file
@@ -0,0 +1,105 @@
#!/usr/bin/env bash

# Copyright   2014  Johns Hopkins University (author: Daniel Povey)
#             2017  Xingyu Na
# Apache 2.0

remove_archive=false

if [ "$1" == --remove-archive ]; then
  remove_archive=true
  shift
fi

if [ $# -ne 3 ]; then
  echo "Usage: $0 [--remove-archive] <data-base> <url-base> <corpus-part>"
  echo "e.g.: $0 /export/a05/xna/data www.openslr.org/resources/33 data_aishell"
  echo "With --remove-archive it will remove the archive after successfully un-tarring it."
  echo "<corpus-part> can be one of: data_aishell, resource_aishell."
fi

data=$1
url=$2
part=$3

if [ ! -d "$data" ]; then
  echo "$0: no such directory $data"
  exit 1;
fi

part_ok=false
list="data_aishell resource_aishell"
for x in $list; do
  if [ "$part" == $x ]; then part_ok=true; fi
done
if ! $part_ok; then
  echo "$0: expected <corpus-part> to be one of $list, but got '$part'"
  exit 1;
fi

if [ -z "$url" ]; then
  echo "$0: empty URL base."
  exit 1;
fi

if [ -f $data/$part/.complete ]; then
  echo "$0: data part $part was already successfully extracted, nothing to do."
  exit 0;
fi

# sizes of the archive files in bytes.
sizes="15582913665 1246920"

if [ -f $data/$part.tgz ]; then
  size=$(/bin/ls -l $data/$part.tgz | awk '{print $5}')
  size_ok=false
  for s in $sizes; do if [ $s == $size ]; then size_ok=true; fi; done
  if ! $size_ok; then
    echo "$0: removing existing file $data/$part.tgz because its size in bytes $size"
    echo "does not equal the size of one of the archives."
    rm $data/$part.tgz
  else
    echo "$data/$part.tgz exists and appears to be complete."
  fi
fi

if [ ! -f $data/$part.tgz ]; then
  if ! command -v wget >/dev/null; then
    echo "$0: wget is not installed."
    exit 1;
  fi
  full_url=$url/$part.tgz
  echo "$0: downloading data from $full_url.  This may take some time, please be patient."

  cd $data || exit 1
  if ! wget --no-check-certificate $full_url; then
    echo "$0: error executing wget $full_url"
    exit 1;
  fi
fi

cd $data || exit 1

if ! tar -xvzf $part.tgz; then
  echo "$0: error un-tarring archive $data/$part.tgz"
  exit 1;
fi

touch $data/$part/.complete

if [ $part == "data_aishell" ]; then
  cd $data/$part/wav || exit 1
  for wav in ./*.tar.gz; do
    echo "Extracting wav from $wav"
    tar -zxf $wav && rm $wav
  done
fi

echo "$0: Successfully downloaded and un-tarred $data/$part.tgz"

if $remove_archive; then
  echo "$0: removing $data/$part.tgz file since --remove-archive option was supplied."
  rm $data/$part.tgz
fi

exit 0;

 egs/aishell/data2vec_paraformer_finetune/local/prepare_data.sh

File was deleted

 egs/aishell/data2vec_transformer_finetune/conf/train_asr_transformer_12e_6d_3072_768.yaml

@@ -30,25 +30,28 @@
  require_same_masks: true
  mask_dropout: 0

# frontend related
frontend: wav_frontend
frontend_conf:
    fs: 16000
    window: hamming
    n_mels: 80
    frame_length: 25
    frame_shift: 10
    lfr_m: 1
    lfr_n: 1

# hybrid CTC/attention
model_conf:
    ctc_weight: 1.0
    lsm_weight: 0.1     # label smoothing option
    length_normalized_loss: false

# for logger
log_interval: 50

# minibatch related
batch_type: length
batch_bins: 16000
num_workers: 16

# optimization related
accum_grad: 1
grad_clip: 5
patience: none
max_epoch: 50
max_epoch: 150
val_scheduler_criterion:
    - valid
    - acc
@@ -57,8 +60,6 @@
    - cer_ctc
    - min
keep_nbest_models: 10
unused_parameters: true
normalize: None

# NoamLR is deprecated. Use WarmupLR.
# The following is equivalent setting for NoamLR:
@@ -92,4 +93,18 @@
    time_mask_width_range:
    - 0
    - 40
    num_time_mask: 2
    num_time_mask: 2

dataset_conf:
    shuffle: True
    shuffle_conf:
        shuffle_size: 2048
        sort_size: 500
    batch_conf:
        batch_type: token
        batch_size: 25000
    num_workers: 8

log_interval: 50
unused_parameters: true
normalize: None

 egs/aishell/data2vec_transformer_finetune/local/download_and_untar.sh

New file
@@ -0,0 +1,105 @@
#!/usr/bin/env bash

# Copyright   2014  Johns Hopkins University (author: Daniel Povey)
#             2017  Xingyu Na
# Apache 2.0

remove_archive=false

if [ "$1" == --remove-archive ]; then
  remove_archive=true
  shift
fi

if [ $# -ne 3 ]; then
  echo "Usage: $0 [--remove-archive] <data-base> <url-base> <corpus-part>"
  echo "e.g.: $0 /export/a05/xna/data www.openslr.org/resources/33 data_aishell"
  echo "With --remove-archive it will remove the archive after successfully un-tarring it."
  echo "<corpus-part> can be one of: data_aishell, resource_aishell."
fi

data=$1
url=$2
part=$3

if [ ! -d "$data" ]; then
  echo "$0: no such directory $data"
  exit 1;
fi

part_ok=false
list="data_aishell resource_aishell"
for x in $list; do
  if [ "$part" == $x ]; then part_ok=true; fi
done
if ! $part_ok; then
  echo "$0: expected <corpus-part> to be one of $list, but got '$part'"
  exit 1;
fi

if [ -z "$url" ]; then
  echo "$0: empty URL base."
  exit 1;
fi

if [ -f $data/$part/.complete ]; then
  echo "$0: data part $part was already successfully extracted, nothing to do."
  exit 0;
fi

# sizes of the archive files in bytes.
sizes="15582913665 1246920"

if [ -f $data/$part.tgz ]; then
  size=$(/bin/ls -l $data/$part.tgz | awk '{print $5}')
  size_ok=false
  for s in $sizes; do if [ $s == $size ]; then size_ok=true; fi; done
  if ! $size_ok; then
    echo "$0: removing existing file $data/$part.tgz because its size in bytes $size"
    echo "does not equal the size of one of the archives."
    rm $data/$part.tgz
  else
    echo "$data/$part.tgz exists and appears to be complete."
  fi
fi

if [ ! -f $data/$part.tgz ]; then
  if ! command -v wget >/dev/null; then
    echo "$0: wget is not installed."
    exit 1;
  fi
  full_url=$url/$part.tgz
  echo "$0: downloading data from $full_url.  This may take some time, please be patient."

  cd $data || exit 1
  if ! wget --no-check-certificate $full_url; then
    echo "$0: error executing wget $full_url"
    exit 1;
  fi
fi

cd $data || exit 1

if ! tar -xvzf $part.tgz; then
  echo "$0: error un-tarring archive $data/$part.tgz"
  exit 1;
fi

touch $data/$part/.complete

if [ $part == "data_aishell" ]; then
  cd $data/$part/wav || exit 1
  for wav in ./*.tar.gz; do
    echo "Extracting wav from $wav"
    tar -zxf $wav && rm $wav
  done
fi

echo "$0: Successfully downloaded and un-tarred $data/$part.tgz"

if $remove_archive; then
  echo "$0: removing $data/$part.tgz file since --remove-archive option was supplied."
  rm $data/$part.tgz
fi

exit 0;

 egs/aishell/data2vec_transformer_finetune/local/prepare_data.sh

File was deleted

 egs/aishell/data2vec_transformer_finetune/run.sh

@@ -8,33 +8,30 @@
count=1
gpu_inference=true  # Whether to perform gpu decoding, set false for cpu decoding
# for gpu decoding, inference_nj=ngpu*njob; for cpu decoding, inference_nj=njob
njob=5
njob=1
train_cmd=utils/run.pl
infer_cmd=utils/run.pl

# general configuration
feats_dir="../DATA" #feature output dictionary, for large data
feats_dir="../DATA" #feature output dictionary
exp_dir="."
lang=zh
dumpdir=dump/fbank
feats_type=fbank
token_type=char
scp=feats.scp
type=kaldi_ark
stage=0
type=sound
scp=wav.scp
stage=3
stop_stage=4

# feature configuration
feats_dim=80
sample_frequency=16000
nj=32
speed_perturb="0.9,1.0,1.1"
nj=64

# data
data_aishell=
raw_data=
data_url=www.openslr.org/resources/33

# exp tag
tag=""
tag="exp1"

model_name=damo/speech_data2vec_pretrain-zh-cn-aishell2-16k-pytorch
init_param="$HOME/.cache/modelscope/hub/$model_name/basemodel.pb"
@@ -52,10 +49,10 @@
test_sets="dev test"

asr_config=conf/train_asr_transformer_12e_6d_3072_768.yaml
model_dir="baseline_$(basename "${asr_config}" .yaml)_${feats_type}_${lang}_${token_type}_${tag}"
model_dir="baseline_$(basename "${asr_config}" .yaml)_${lang}_${token_type}_${tag}"

inference_config=conf/decode_asr_transformer.yaml
inference_asr_model=valid.cer_ctc.ave_10best.pb
inference_config=conf/decode_asr_transformer_noctc_1best.yaml
inference_asr_model=valid.acc.ave_10best.pb

# you can set gpu num for decoding here
gpuid_list=$CUDA_VISIBLE_DEVICES  # set gpus for decoding, the same as training stage by default
@@ -69,10 +66,16 @@
    _ngpu=0
fi

if [ ${stage} -le -1 ] && [ ${stop_stage} -ge -1 ]; then
    echo "stage -1: Data Download"
    local/download_and_untar.sh ${raw_data} ${data_url} data_aishell
    local/download_and_untar.sh ${raw_data} ${data_url} resource_aishell
fi

if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
    echo "stage 0: Data preparation"
    # Data preparation
    local/aishell_data_prep.sh ${data_aishell}/data_aishell/wav ${data_aishell}/data_aishell/transcript ${feats_dir}
    local/aishell_data_prep.sh ${raw_data}/data_aishell/wav ${raw_data}/data_aishell/transcript ${feats_dir}
    for x in train dev test; do
        cp ${feats_dir}/data/${x}/text ${feats_dir}/data/${x}/text.org
        paste -d " " <(cut -f 1 -d" " ${feats_dir}/data/${x}/text.org) <(cut -f 2- -d" " ${feats_dir}/data/${x}/text.org | tr -d " ") \
@@ -82,46 +85,9 @@
    done
fi

feat_train_dir=${feats_dir}/${dumpdir}/train; mkdir -p ${feat_train_dir}
feat_dev_dir=${feats_dir}/${dumpdir}/dev; mkdir -p ${feat_dev_dir}
feat_test_dir=${feats_dir}/${dumpdir}/test; mkdir -p ${feat_test_dir}
if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
    echo "stage 1: Feature Generation"
    # compute fbank features
    fbankdir=${feats_dir}/fbank
    utils/compute_fbank.sh --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} --sample_frequency ${sample_frequency} --speed_perturb ${speed_perturb} \
        ${feats_dir}/data/train ${exp_dir}/exp/make_fbank/train ${fbankdir}/train
    utils/fix_data_feat.sh ${fbankdir}/train
    utils/compute_fbank.sh --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} --sample_frequency ${sample_frequency} \
        ${feats_dir}/data/dev ${exp_dir}/exp/make_fbank/dev ${fbankdir}/dev
    utils/fix_data_feat.sh ${fbankdir}/dev
    utils/compute_fbank.sh --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} --sample_frequency ${sample_frequency} \
        ${feats_dir}/data/test ${exp_dir}/exp/make_fbank/test ${fbankdir}/test
    utils/fix_data_feat.sh ${fbankdir}/test
     
    # compute global cmvn
    utils/compute_cmvn.sh --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} \
        ${fbankdir}/train ${exp_dir}/exp/make_fbank/train

    # apply cmvn 
    utils/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
        ${fbankdir}/train ${fbankdir}/train/cmvn.json ${exp_dir}/exp/make_fbank/train ${feat_train_dir}
    utils/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
        ${fbankdir}/dev ${fbankdir}/train/cmvn.json ${exp_dir}/exp/make_fbank/dev ${feat_dev_dir}
    utils/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
        ${fbankdir}/test ${fbankdir}/train/cmvn.json ${exp_dir}/exp/make_fbank/test ${feat_test_dir}
    
    cp ${fbankdir}/train/text ${fbankdir}/train/speech_shape ${fbankdir}/train/text_shape ${feat_train_dir}
    cp ${fbankdir}/dev/text ${fbankdir}/dev/speech_shape ${fbankdir}/dev/text_shape ${feat_dev_dir}
    cp ${fbankdir}/test/text ${fbankdir}/test/speech_shape ${fbankdir}/test/text_shape ${feat_test_dir}

    utils/fix_data_feat.sh ${feat_train_dir}
    utils/fix_data_feat.sh ${feat_dev_dir}
    utils/fix_data_feat.sh ${feat_test_dir}

    #generate ark list 
    utils/gen_ark_list.sh --cmd "$train_cmd" --nj $nj ${feat_train_dir} ${fbankdir}/train ${feat_train_dir}
    utils/gen_ark_list.sh --cmd "$train_cmd" --nj $nj ${feat_dev_dir} ${fbankdir}/dev ${feat_dev_dir}
    echo "stage 1: Feature and CMVN Generation"
    utils/compute_cmvn.sh --cmd "$train_cmd" --nj $nj --feats_dim ${feats_dim} ${feats_dir}/data/${train_set}
fi

token_list=${feats_dir}/data/${lang}_token_list/char/tokens.txt
@@ -129,35 +95,27 @@
if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
    echo "stage 2: Dictionary Preparation"
    mkdir -p ${feats_dir}/data/${lang}_token_list/char/
   

    echo "make a dictionary"
    echo "<blank>" > ${token_list}
    echo "<s>" >> ${token_list}
    echo "</s>" >> ${token_list}
    utils/text2token.py -s 1 -n 1 --space "" ${feats_dir}/data/train/text | cut -f 2- -d" " | tr " " "\n" \
    utils/text2token.py -s 1 -n 1 --space "" ${feats_dir}/data/$train_set/text | cut -f 2- -d" " | tr " " "\n" \
        | sort | uniq | grep -a -v -e '^\s*$' | awk '{print $0}' >> ${token_list}
    num_token=$(cat ${token_list} | wc -l)
    echo "<unk>" >> ${token_list}
    vocab_size=$(cat ${token_list} | wc -l)
    awk -v v=,${vocab_size} '{print $0v}' ${feat_train_dir}/text_shape > ${feat_train_dir}/text_shape.char
    awk -v v=,${vocab_size} '{print $0v}' ${feat_dev_dir}/text_shape > ${feat_dev_dir}/text_shape.char
    mkdir -p ${feats_dir}/asr_stats_fbank_zh_char/train 
    mkdir -p ${feats_dir}/asr_stats_fbank_zh_char/dev
    cp ${feat_train_dir}/speech_shape ${feat_train_dir}/text_shape ${feat_train_dir}/text_shape.char ${feats_dir}/asr_stats_fbank_zh_char/train
    cp ${feat_dev_dir}/speech_shape ${feat_dev_dir}/text_shape ${feat_dev_dir}/text_shape.char ${feats_dir}/asr_stats_fbank_zh_char/dev
fi

# Training Stage
world_size=$gpu_num  # run on one machine
if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
    echo "stage 3: Training"
    python utils/download_model.py  --model_name ${model_name}  # download pretrained model on ModelScope
     python utils/download_model.py  --model_name ${model_name}  # download pretrained model on ModelScope
    mkdir -p ${exp_dir}/exp/${model_dir}
    mkdir -p ${exp_dir}/exp/${model_dir}/log
    INIT_FILE=${exp_dir}/exp/${model_dir}/ddp_init
    if [ -f $INIT_FILE ];then
        rm -f $INIT_FILE
    fi 
    fi
    init_method=file://$(readlink -f $INIT_FILE)
    echo "$0: init method is $init_method"
    for ((i = 0; i < $gpu_num; ++i)); do
@@ -165,27 +123,22 @@
            rank=$i
            local_rank=$i
            gpu_id=$(echo $CUDA_VISIBLE_DEVICES | cut -d',' -f$[$i+1])
            asr_train.py \
            train.py \
                --task_name asr \
                --gpu_id $gpu_id \
                --use_preprocessor true \
                --token_type char \
                --token_list $token_list \
                --train_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${train_set}/${scp},speech,${type} \
                --train_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${train_set}/text,text,text \
                --train_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${train_set}/speech_shape \
                --train_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${train_set}/text_shape.char \
                --valid_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${valid_set}/${scp},speech,${type} \
                --valid_data_path_and_name_and_type ${feats_dir}/${dumpdir}/${valid_set}/text,text,text \
                --valid_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${valid_set}/speech_shape \
                --valid_shape_file ${feats_dir}/asr_stats_fbank_zh_char/${valid_set}/text_shape.char  \
                --data_dir ${feats_dir}/data \
                --train_set ${train_set} \
                --valid_set ${valid_set} \
                --init_param ${init_param} \
                --cmvn_file ${feats_dir}/data/${train_set}/cmvn/cmvn.mvn \
                --resume true \
                --output_dir ${exp_dir}/exp/${model_dir} \
                --config $asr_config \
                --input_size $feats_dim \
                --ngpu $gpu_num \
                --num_worker_count $count \
                --multiprocessing_distributed true \
                --dist_init_method $init_method \
                --dist_world_size $world_size \
                --dist_rank $rank \
@@ -208,7 +161,7 @@
            exit 0
        fi
        mkdir -p "${_logdir}"
        _data="${feats_dir}/${dumpdir}/${dset}"
        _data="${feats_dir}/data/${dset}"
        key_file=${_data}/${scp}
        num_scp_file="$(<${key_file} wc -l)"
        _nj=$([ $inference_nj -le $num_scp_file ] && echo "$inference_nj" || echo "$num_scp_file")
@@ -229,11 +182,12 @@
                --njob ${njob} \
                --gpuid_list ${gpuid_list} \
                --data_path_and_name_and_type "${_data}/${scp},speech,${type}" \
                --cmvn_file ${feats_dir}/data/${train_set}/cmvn/cmvn.mvn \
                --key_file "${_logdir}"/keys.JOB.scp \
                --asr_train_config "${asr_exp}"/config.yaml \
                --asr_model_file "${asr_exp}"/"${inference_asr_model}" \
                --output_dir "${_logdir}"/output.JOB \
                --mode asr \
                --mode paraformer \
                ${_opts}

        for f in token token_int score text; do
@@ -249,4 +203,4 @@
        tail -n 3 ${_dir}/text.cer > ${_dir}/text.cer.txt
        cat ${_dir}/text.cer.txt
    done
fi
fi