clean prepare.sh

egs/mucs/ASR/prepare.sh

@@ -6,36 +6,15 @@ export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
 set -eou pipefail
 
 nj=60
-stage=-1
+stage=6
 stop_stage=9
 
 # We assume dl_dir (download dir) contains the following
 # directories and files. If not, they will be downloaded
 # by this script automatically.
 #
-#  - $dl_dir/LibriSpeech
+#  - $dl_dir/hi-en
-#      You can find BOOKS.TXT, test-clean, train-clean-360, etc, inside it.
+
-#      You can download them from https://www.openslr.org/12
-#
-#  - $dl_dir/lm
-#      This directory contains the following files downloaded from
-#       http://www.openslr.org/resources/11
-#
-#        - 3-gram.pruned.1e-7.arpa.gz
-#        - 3-gram.pruned.1e-7.arpa
-#        - 4-gram.arpa.gz
-#        - 4-gram.arpa
-#        - librispeech-vocab.txt
-#        - librispeech-lexicon.txt
-#        - librispeech-lm-norm.txt.gz
-#
-#  - $dl_dir/musan
-#      This directory contains the following directories downloaded from
-#       http://www.openslr.org/17/
-#
-#     - music
-#     - noise
-#     - speech
 dl_dir=$PWD/download
 espnet_path=/home/wtc7/espnet/egs2/MUCS/asr1/data/hi-en/
 
@@ -43,13 +22,8 @@ espnet_path=/home/wtc7/espnet/egs2/MUCS/asr1/data/hi-en/
 
 # vocab size for sentence piece models.
 # It will generate data/lang_bpe_xxx,
-# data/lang_bpe_yyy if the array contains xxx, yyy
+# data/lang_bpe_yyy
-vocab_sizes=(
+vocab_size=400
-  # 5000
-  # 2000
-  # 1000
-  200
-)
 
 # All files generated by this script are saved in "data".
 # You can safely remove "data" and rerun this script to regenerate it.
@@ -68,7 +42,7 @@ if [ $stage -le -1 ] && [ $stop_stage -ge -1 ]; then
   mkdir -p $dl_dir/lm
   if [ ! -e $dl_dir/lm/.done ]; then
     ./local/prepare_lm_files.py --out-dir=$dl_dir/lm --data-path=$espnet_path --mode="train"
-    # touch $dl_dir/lm/.done
+    touch $dl_dir/lm/.done
   fi
 fi
 
@@ -78,11 +52,11 @@ fi
 
 if [ $stage -le 1 ] && [ $stop_stage -ge 1 ]; then
   log "Stage 1: Prepare MUCS manifest"
-  # We assume that you have downloaded the LibriSpeech corpus
+  # We assume that you have downloaded the MUCS corpus
-  # to $dl_dir/LibriSpeech
+  # to $dl_dir/
   mkdir -p data/manifests
   if [ ! -e data/manifests/.mucs.done ]; then
-    # lhotse prepare mucs -j $nj $dl_dir/hi-en data/manifests
+    # generate lhotse manifests from kaldi style files
     ./local/prepare_manifest.py "$espnet_path" $nj data/manifests
 
     touch data/manifests/.mucs.done
@@ -94,7 +68,7 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
   mkdir -p data/fbank
   if [ ! -e data/fbank/.mucs.done ]; then
     ./local/compute_fbank_mucs.py
-    # touch data/fbank/.mucs.done
+    touch data/fbank/.mucs.done
   fi
 
   # exit
@@ -110,7 +84,7 @@ if [ $stage -le 3 ] && [ $stop_stage -ge 3 ]; then
       python3 ./local/validate_manifest.py \
         data/fbank/mucs_cuts_${part}.jsonl.gz
     done
-    # touch data/fbank/.mucs-validated.done
+    touch data/fbank/.mucs-validated.done
   fi
 fi
 
@@ -150,7 +124,6 @@ fi
 if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
   log "Stage 6: Prepare BPE based lang"
 
-  for vocab_size in ${vocab_sizes[@]}; do
     lang_dir=data/lang_bpe_${vocab_size}
     mkdir -p $lang_dir
     # We reuse words.txt from phone based lexicon
@@ -193,23 +166,14 @@ if [ $stage -le 6 ] && [ $stop_stage -ge 6 ]; then
         $lang_dir/L_disambig.pt \
         $lang_dir/L_disambig.fst
     fi
-  done
+  
 fi
 
 if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
-  log "Stage 7: Prepare bigram token-level P for MMI training"
+  log "Stage 7: Train LM from training data"
 
-  for vocab_size in ${vocab_sizes[@]}; do
     lang_dir=data/lang_bpe_${vocab_size}
 
-    # if [ ! -f $lang_dir/transcript_tokens.txt ]; then
-    #   ./local/convert_transcript_words_to_tokens.py \
-    #     --lexicon $lang_dir/lexicon.txt \
-    #     --transcript $lang_dir/transcript_words.txt \
-    #     --oov "<UNK>" \
-    #     > $lang_dir/transcript_tokens.txt
-    # fi
-
     if [ ! -f $lang_dir/lm_3.arpa ]; then
       ./shared/make_kn_lm.py \
         -ngram-order 3 \
@@ -224,14 +188,6 @@ if [ $stage -le 7 ] && [ $stop_stage -ge 7 ]; then
         -lm $lang_dir/lm_4.arpa
     fi
 
-    # if [ ! -f $lang_dir/P.fst.txt ]; then
-    #   python3 -m kaldilm \
-    #     --read-symbol-table="$lang_dir/tokens.txt" \
-    #     --disambig-symbol='#0' \
-    #     --max-order=2 \
-    #     $lang_dir/P.arpa > $lang_dir/P.fst.txt
-    # fi
-  done
 fi
 
 if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
@@ -246,7 +202,7 @@ if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
       --read-symbol-table="data/lang_phone/words.txt" \
       --disambig-symbol='#0' \
       --max-order=3 \
-      data/lang_bpe_200/lm_3.arpa > data/lm/G_3_gram.fst.txt
+      data/lang_bpe_${vocab_size}/lm_3.arpa > data/lm/G_3_gram.fst.txt
   fi
 
   if [ ! -f data/lm/G_4_gram.fst.txt ]; then
@@ -255,17 +211,9 @@ if [ $stage -le 8 ] && [ $stop_stage -ge 8 ]; then
       --read-symbol-table="data/lang_phone/words.txt" \
       --disambig-symbol='#0' \
       --max-order=3 \
-      data/lang_bpe_200/lm_4.arpa > data/lm/G_4_gram.fst.txt
+      data/lang_bpe_${vocab_size}/lm_4.arpa > data/lm/G_4_gram.fst.txt
   fi
 
-  # if [ ! -f data/lm/G_4_gram.fst.txt ]; then
-  #   # It is used for LM rescoring
-  #   python3 -m kaldilm \
-  #     --read-symbol-table="data/lang_phone/words.txt" \
-  #     --disambig-symbol='#0' \
-  #     --max-order=4 \
-  #     $dl_dir/lm/4-gram.arpa > data/lm/G_4_gram.fst.txt
-  # fi
 fi
 
 if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
@@ -277,120 +225,8 @@ if [ $stage -le 9 ] && [ $stop_stage -ge 9 ]; then
   #
   # ./local/compile_hlg_using_openfst.py --lang-dir data/lang_phone
 
-  for vocab_size in ${vocab_sizes[@]}; do
     lang_dir=data/lang_bpe_${vocab_size}
     ./local/compile_hlg.py --lang-dir $lang_dir
 
-    # Note If ./local/compile_hlg.py throws OOM,
-    # please switch to the following command
-    #
-    # ./local/compile_hlg_using_openfst.py --lang-dir $lang_dir
-  done
 fi
 
-# Compile LG for RNN-T fast_beam_search decoding
-if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
-  log "Stage 10: Compile LG"
-  ./local/compile_lg.py --lang-dir data/lang_phone
-
-  for vocab_size in ${vocab_sizes[@]}; do
-    lang_dir=data/lang_bpe_${vocab_size}
-    ./local/compile_lg.py --lang-dir $lang_dir
-  done
-fi
-
-if [ $stage -le 11 ] && [ $stop_stage -ge 11 ]; then
-  log "Stage 11: Generate LM training data"
-
-  for vocab_size in ${vocab_sizes[@]}; do
-    log "Processing vocab_size == ${vocab_size}"
-    lang_dir=data/lang_bpe_${vocab_size}
-    out_dir=data/lm_training_bpe_${vocab_size}
-    mkdir -p $out_dir
-
-    ./local/prepare_lm_training_data.py \
-      --bpe-model $lang_dir/bpe.model \
-      --lm-data $dl_dir/lm/librispeech-lm-norm.txt \
-      --lm-archive $out_dir/lm_data.pt
-  done
-fi
-
-if [ $stage -le 12 ] && [ $stop_stage -ge 12 ]; then
-  log "Stage 12: Generate LM validation data"
-
-  for vocab_size in ${vocab_sizes[@]}; do
-    log "Processing vocab_size == ${vocab_size}"
-    out_dir=data/lm_training_bpe_${vocab_size}
-    mkdir -p $out_dir
-
-    if [ ! -f $out_dir/valid.txt ]; then
-      files=$(
-        find "$dl_dir/LibriSpeech/dev-clean" -name "*.trans.txt"
-        find "$dl_dir/LibriSpeech/dev-other" -name "*.trans.txt"
-      )
-      for f in ${files[@]}; do
-        cat $f | cut -d " " -f 2-
-      done > $out_dir/valid.txt
-    fi
-
-    lang_dir=data/lang_bpe_${vocab_size}
-    ./local/prepare_lm_training_data.py \
-      --bpe-model $lang_dir/bpe.model \
-      --lm-data $out_dir/valid.txt \
-      --lm-archive $out_dir/lm_data-valid.pt
-  done
-fi
-
-if [ $stage -le 13 ] && [ $stop_stage -ge 13 ]; then
-  log "Stage 13: Generate LM test data"
-
-  for vocab_size in ${vocab_sizes[@]}; do
-    log "Processing vocab_size == ${vocab_size}"
-    out_dir=data/lm_training_bpe_${vocab_size}
-    mkdir -p $out_dir
-
-    if [ ! -f $out_dir/test.txt ]; then
-      files=$(
-        find "$dl_dir/LibriSpeech/test-clean" -name "*.trans.txt"
-        find "$dl_dir/LibriSpeech/test-other" -name "*.trans.txt"
-      )
-      for f in ${files[@]}; do
-        cat $f | cut -d " " -f 2-
-      done > $out_dir/test.txt
-    fi
-
-    lang_dir=data/lang_bpe_${vocab_size}
-    ./local/prepare_lm_training_data.py \
-      --bpe-model $lang_dir/bpe.model \
-      --lm-data $out_dir/test.txt \
-      --lm-archive $out_dir/lm_data-test.pt
-  done
-fi
-
-if [ $stage -le 14 ] && [ $stop_stage -ge 14 ]; then
-  log "Stage 14: Sort LM training data"
-  # Sort LM training data by sentence length in descending order
-  # for ease of training.
-  #
-  # Sentence length equals to the number of BPE tokens
-  # in a sentence.
-
-  for vocab_size in ${vocab_sizes[@]}; do
-    out_dir=data/lm_training_bpe_${vocab_size}
-    mkdir -p $out_dir
-    ./local/sort_lm_training_data.py \
-      --in-lm-data $out_dir/lm_data.pt \
-      --out-lm-data $out_dir/sorted_lm_data.pt \
-      --out-statistics $out_dir/statistics.txt
-
-    ./local/sort_lm_training_data.py \
-      --in-lm-data $out_dir/lm_data-valid.pt \
-      --out-lm-data $out_dir/sorted_lm_data-valid.pt \
-      --out-statistics $out_dir/statistics-valid.txt
-
-    ./local/sort_lm_training_data.py \
-      --in-lm-data $out_dir/lm_data-test.pt \
-      --out-lm-data $out_dir/sorted_lm_data-test.pt \
-      --out-statistics $out_dir/statistics-test.txt
-  done
-fi