update repo

egs/aishell/paraformerbert/run.sh

@@ -111,12 +111,12 @@ fi
 world_size=$gpu_num  # run on one machine
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
     echo "stage 3: Training"
-#    if ! "${skip_extract_embed}"; then
-#        echo "extract embeddings..."
-#        local/extract_embeds.sh \
-#            --bert_model_name ${bert_model_name} \
-#            --raw_dataset_path ${feats_dir}
-#    fi
+    if ! "${skip_extract_embed}"; then
+        echo "extract embeddings..."
+        local/extract_embeds.sh \
+            --bert_model_name ${bert_model_name} \
+            --raw_dataset_path ${feats_dir}
+    fi
     mkdir -p ${exp_dir}/exp/${model_dir}
     mkdir -p ${exp_dir}/exp/${model_dir}/log
     INIT_FILE=${exp_dir}/exp/${model_dir}/ddp_init

egs/aishell2/paraformerbert/local/extract_embeds.sh

@@ -3,20 +3,17 @@
 stage=1
 stop_stage=3
 
-bert_model_root="../../huggingface_models"
 bert_model_name="bert-base-chinese"
-#bert_model_name="chinese-roberta-wwm-ext"
-#bert_model_name="mengzi-bert-base"
 raw_dataset_path="../DATA"
-model_path=${bert_model_root}/${bert_model_name}
+model_path=${bert_model_name}
 
 . utils/parse_options.sh || exit 1;
 
-nj=100
+nj=32
 
-for data_set in train dev_ios test_ios;do
-    scp=$raw_dataset_path/dump/fbank/${data_set}/text
-    local_scp_dir_raw=$raw_dataset_path/embeds/$bert_model_name/${data_set}
+for data_set in train dev test;do
+    scp=$raw_dataset_path/data/${data_set}/text
+    local_scp_dir_raw=${raw_dataset_path}/data/embeds/${data_set}
     local_scp_dir=$local_scp_dir_raw/split$nj
     local_records_dir=$local_scp_dir_raw/ark
 
@@ -31,7 +28,7 @@ for data_set in train dev_ios test_ios;do
     utils/split_scp.pl $scp ${split_scps}
 
 
-    for num in {0..24};do
+    for num in {0..7};do
         tmp=`expr $num \* 4`
 
         if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
@@ -41,23 +38,12 @@ for data_set in train dev_ios test_ios;do
                 {
                     beg=0
                     gpu=`expr $beg + $idx`
-                    echo $local_scp_dir_raw/log/log.${JOB}
-                    python tools/extract_embeds.py $local_scp_dir/text.$JOB.txt ${local_records_dir}/embeds.${JOB}.ark ${local_records_dir}/embeds.${JOB}.scp ${local_records_dir}/embeds.${JOB}.shape ${gpu} ${model_path} &> $local_scp_dir_raw/log/log.${JOB}
+                    echo ${local_scp_dir}/log.${JOB}
+                    python utils/extract_embeds.py $local_scp_dir/data.$JOB.text ${local_records_dir}/embeds.${JOB}.ark ${local_records_dir}/embeds.${JOB}.scp ${local_records_dir}/embeds.${JOB}.shape ${gpu} ${model_path} &> ${local_scp_dir}/log.${JOB}
             } &
             done
             wait
         fi
-
-        if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
-            for idx in {1..4}; do
-                JOB=`expr $tmp + $idx`
-                echo "upload jobid=$JOB"
-                {
-                    hadoop  fs -put -f ${local_records_dir}/embeds.${JOB}.ark ${odps_des_feature_dir}/embeds.${JOB}.ark
-                } &
-            done
-            wait
-        fi
     done
 
     if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then

egs/aishell2/paraformerbert/local/prepare_data.sh

@@ -17,7 +17,6 @@ if [ $# != 3 ]; then
 fi
 
 corpus=$1
-#dict_dir=$2
 tmp=$2
 dir=$3
 
@@ -35,14 +34,14 @@ fi
 # validate utt-key list, IC0803W0380 is a bad utterance
 awk '{print $1}' $corpus/wav.scp | grep -v 'IC0803W0380' > $tmp/wav_utt.list
 awk '{print $1}' $corpus/trans.txt > $tmp/trans_utt.list
-tools/filter_scp.pl -f 1 $tmp/wav_utt.list $tmp/trans_utt.list > $tmp/utt.list
+utils/filter_scp.pl -f 1 $tmp/wav_utt.list $tmp/trans_utt.list > $tmp/utt.list
 
 # wav.scp
 awk -F'\t' -v path_prefix=$corpus '{printf("%s\t%s/%s\n",$1,path_prefix,$2)}' $corpus/wav.scp > $tmp/tmp_wav.scp
-tools/filter_scp.pl -f 1 $tmp/utt.list $tmp/tmp_wav.scp | sort -k 1 | uniq > $tmp/wav.scp
+utils/filter_scp.pl -f 1 $tmp/utt.list $tmp/tmp_wav.scp | sort -k 1 | uniq > $tmp/wav.scp
 
 # text
-tools/filter_scp.pl -f 1 $tmp/utt.list $corpus/trans.txt | sort -k 1 | uniq > $tmp/text
+utils/filter_scp.pl -f 1 $tmp/utt.list $corpus/trans.txt | sort -k 1 | uniq > $tmp/text
 
 # copy prepared resources from tmp_dir to target dir
 mkdir -p $dir

egs/aishell2/paraformerbert/run.sh

@@ -8,36 +8,32 @@ gpu_num=8
 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
-train_cmd=tools/run.pl
+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
+type=sound
+scp=wav.scp
+speed_perturb="0.9 1.0 1.1"
 dataset_type=large
-scp=feats.scp
-type=kaldi_ark
-stage=0
-stop_stage=5
+stage=3
+stop_stage=4
 
 skip_extract_embed=false
-bert_model_root="../../huggingface_models"
 bert_model_name="bert-base-chinese"
 
 # feature configuration
 feats_dim=80
-sample_frequency=16000
-nj=100
-speed_perturb="0.9,1.0,1.1"
+nj=64
 
 # data
-tr_dir=
-dev_tst_dir=
+tr_dir=/nfs/wangjiaming.wjm/asr_data/aishell2/AISHELL-2/iOS/data
+dev_tst_dir=/nfs/wangjiaming.wjm/asr_data/aishell2/AISHELL-DEV-TEST-SET
 
 # exp tag
 tag="exp1"
@@ -55,7 +51,7 @@ valid_set=dev_ios
 test_sets="dev_ios test_ios"
 
 asr_config=conf/train_asr_paraformerbert_conformer_20e_6d_1280_320.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_noctc_1best.yaml
 inference_asr_model=valid.acc.ave_10best.pb
@@ -75,86 +71,44 @@ fi
 if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ]; then
     echo "stage 0: Data preparation"
     # For training set
-    local/prepare_data.sh ${tr_dir} data/local/train data/train || exit 1;
+    local/prepare_data.sh ${tr_dir} ${feats_dir}/data/local/train ${feats_dir}/data/train || exit 1;
     # # For dev and test set
-    for x in Android iOS Mic; do
-        local/prepare_data.sh ${dev_tst_dir}/${x}/dev data/local/dev_${x,,} data/dev_${x,,} || exit 1;
-        local/prepare_data.sh ${dev_tst_dir}/${x}/test data/local/test_${x,,} data/test_${x,,} || exit 1;
-    done 
+    for x in iOS; do
+        local/prepare_data.sh ${dev_tst_dir}/${x}/dev ${feats_dir}/data/local/dev_${x,,} ${feats_dir}/data/dev_${x,,} || exit 1;
+        local/prepare_data.sh ${dev_tst_dir}/${x}/test ${feats_dir}/data/local/test_${x,,} ${feats_dir}/data/test_${x,,} || exit 1;
+    done
     # Normalize text to capital letters
-    for x in train dev_android dev_ios dev_mic test_android test_ios test_mic; do
-        mv data/${x}/text data/${x}/text.org
-        paste <(cut -f 1 data/${x}/text.org) <(cut -f 2 data/${x}/text.org | tr '[:lower:]' '[:upper:]') \
-            > data/${x}/text
-        tools/text2token.py -n 1 -s 1 data/${x}/text > data/${x}/text.org
-        mv data/${x}/text.org data/${x}/text
+    for x in train dev_ios test_ios; do
+        mv ${feats_dir}/data/${x}/text ${feats_dir}/data/${x}/text.org
+        paste -d " " <(cut -f 1 ${feats_dir}/data/${x}/text.org) <(cut -f 2- ${feats_dir}/data/${x}/text.org \
+             | tr 'A-Z' 'a-z' | tr -d " ") \
+            > ${feats_dir}/data/${x}/text
+        utils/text2token.py -n 1 -s 1 ${feats_dir}/data/${x}/text > ${feats_dir}/data/${x}/text.org
+        mv ${feats_dir}/data/${x}/text.org ${feats_dir}/data/${x}/text
     done
 fi
 
-feat_train_dir=${feats_dir}/${dumpdir}/${train_set}; mkdir -p ${feat_train_dir}
-feat_dev_dir=${feats_dir}/${dumpdir}/${valid_set}; mkdir -p ${feat_dev_dir}
 if [ ${stage} -le 1 ] && [ ${stop_stage} -ge 1 ]; then
-    echo "stage 1: Feature Generation"
-    # compute fbank features
-    fbankdir=${feats_dir}/fbank
-    steps/compute_fbank.sh --cmd "$train_cmd" --nj $nj --speed_perturb ${speed_perturb} \
-        data/train exp/make_fbank/train ${fbankdir}/train
-    tools/fix_data_feat.sh ${fbankdir}/train
-    for x in android ios mic; do
-        steps/compute_fbank.sh --cmd "$train_cmd" --nj $nj \
-            data/dev_${x} exp/make_fbank/dev_${x} ${fbankdir}/dev_${x}
-        tools/fix_data_feat.sh ${fbankdir}/dev_${x}
-        steps/compute_fbank.sh --cmd "$train_cmd" --nj $nj \
-            data/test_${x} exp/make_fbank/test_${x} ${fbankdir}/test_${x}
-        tools/fix_data_feat.sh ${fbankdir}/test_${x}
-    done
-    
-    # compute global cmvn
-    steps/compute_cmvn.sh --cmd "$train_cmd" --nj $nj \
-        ${fbankdir}/train exp/make_fbank/train
-
-    # apply cmvn 
-    steps/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
-        ${fbankdir}/${train_set} ${fbankdir}/train/cmvn.json exp/make_fbank/${train_set} ${feat_train_dir}
-    steps/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
-        ${fbankdir}/${valid_set} ${fbankdir}/train/cmvn.json exp/make_fbank/${valid_set} ${feat_dev_dir}
-    for x in android ios mic; do
-        steps/apply_cmvn.sh --cmd "$train_cmd" --nj $nj \
-            ${fbankdir}/test_${x} ${fbankdir}/train/cmvn.json exp/make_fbank/test_${x} ${feats_dir}/${dumpdir}/test_${x}
-    done
-    
-    cp ${fbankdir}/${train_set}/text ${fbankdir}/${train_set}/speech_shape ${fbankdir}/${train_set}/text_shape ${feat_train_dir}
-    tools/fix_data_feat.sh ${feat_train_dir}
-    cp ${fbankdir}/${valid_set}/text ${fbankdir}/${valid_set}/speech_shape ${fbankdir}/${valid_set}/text_shape ${feat_dev_dir}
-    tools/fix_data_feat.sh ${feat_dev_dir}
-    for x in android ios mic; do
-        cp ${fbankdir}/test_${x}/text ${fbankdir}/test_${x}/speech_shape ${fbankdir}/test_${x}/text_shape ${feats_dir}/${dumpdir}/test_${x}
-        tools/fix_data_feat.sh ${feats_dir}/${dumpdir}/test_${x}
-    done
+    echo "stage 1: Feature and CMVN Generation"
+    utils/compute_cmvn.sh --cmd "$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
 echo "dictionary: ${token_list}"
 if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
     echo "stage 2: Dictionary Preparation"
-    mkdir -p data/${lang}_token_list/char/
-   
+    mkdir -p ${feats_dir}/data/${lang}_token_list/char/
+
     echo "make a dictionary"
     echo "<blank>" > ${token_list}
     echo "<s>" >> ${token_list}
     echo "</s>" >> ${token_list}
-    tools/text2token.py -s 1 -n 1 --space "" data/${train_set}/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 > ${feat_dev_dir}/text_shape.char
-    mkdir -p asr_stats_fbank_zh_char/${train_set} 
-    mkdir -p asr_stats_fbank_zh_char/${valid_set}
-    cp ${feat_train_dir}/speech_shape ${feat_train_dir}/text_shape ${feat_train_dir}/text_shape.char asr_stats_fbank_zh_char/${train_set} 
-    cp ${feat_dev_dir}/speech_shape ${feat_dev_dir}/text_shape ${feat_dev_dir}/text_shape.char asr_stats_fbank_zh_char/${valid_set}
-fi
+    mkdir -p ${feats_dir}/asr_stats_fbank_zh_char/${train_set}
+    mkdir -p ${feats_dir}/asr_stats_fbank_zh_char/${valid_set}
+ fi
 
 # Training Stage
 world_size=$gpu_num  # run on one machine

funasr/datasets/large_datasets/utils/tokenize.py

@@ -37,7 +37,7 @@ def tokenize(data,
     vad = -2
 
     if bpe_tokenizer is not None:
-        text = bpe_tokenizer.text2tokens("".join(text))
+        text = bpe_tokenizer.text2tokens(text)
 
     if seg_dict is not None:
         assert isinstance(seg_dict, dict)