gen lexicon with tone in mfa, test=tts

examples/other/mfa/local/reorganize_baker.py

@@ -42,9 +42,6 @@ def get_transcripts(path: Union[str, Path]):
     for i in range(0, len(lines), 2):
         sentence_id = lines[i].split()[0]
         transcription = lines[i + 1].strip()
-        # tones are dropped here
-        # since the lexicon does not consider tones, too
-        transcription = " ".join([item[:-1] for item in transcription.split()])
         transcripts[sentence_id] = transcription
 
     return transcripts

examples/other/mfa/run.sh

@@ -4,7 +4,7 @@ mkdir -p $EXP_DIR
 LEXICON_NAME='simple'
 if [ ! -f "$EXP_DIR/$LEXICON_NAME.lexicon" ]; then
     echo "generating lexicon..."
-    python local/generate_lexicon.py "$EXP_DIR/$LEXICON_NAME" --with-r
+    python local/generate_lexicon.py "$EXP_DIR/$LEXICON_NAME" --with-r --with-tone
     echo "lexicon done"
 fi
 
@@ -16,6 +16,7 @@ if [ ! -d $EXP_DIR/baker_corpus ]; then
     echo "transcription for each audio file is saved with the same namd in $EXP_DIR/baker_corpus "
 fi
 
+
 echo "detecting oov..."
 python local/detect_oov.py $EXP_DIR/baker_corpus $EXP_DIR/"$LEXICON_NAME.lexicon"
 echo "detecting oov done. you may consider regenerate lexicon if there is unexpected OOVs."
@@ -44,6 +45,3 @@ if [ ! -d "$EXP_DIR/baker_alignment" ]; then
     echo "model: $EXP_DIR/baker_model"
 fi
 
-
-
-