delete symbol.py

.vs/VSWorkspaceState.json

@@ -3,6 +3,6 @@
     "",
     "\\text"
   ],
-  "SelectedNode": "\\text\\cleaners.py",
+  "SelectedNode": "\\train.py",
   "PreviewInSolutionExplorer": false
 }

data_utils.py

@@ -215,7 +215,7 @@ def get_audio(self, filename):
                 self.sampling_rate, self.hop_length, self.win_length,
                 center=False)
             spec = torch.squeeze(spec, 0)
-            torch.save(spec, spec_filename)
+            torch.save(spec, spec_filename, _use_new_zipfile_serialization=False)
         return spec, audio_norm
 
     def get_text(self, text):

text/__init__.py

@@ -1,11 +1,13 @@
 """ from https://github.com/keithito/tacotron """
 from text import cleaners
-from text.symbols import symbols
 
 
-# Mappings from symbol to numeric ID and vice versa:
-_symbol_to_id = {s: i for i, s in enumerate(symbols)}
-_id_to_symbol = {i: s for i, s in enumerate(symbols)}
+def initialize(symbols):
+  # Mappings from symbol to numeric ID and vice versa:
+  global _symbol_to_id
+  global _id_to_symbol
+  _symbol_to_id = {s: i for i, s in enumerate(symbols)}
+  _id_to_symbol = {i: s for i, s in enumerate(symbols)}
 
 
 def text_to_sequence(text, cleaner_names):

text/cleaners.py

@@ -295,44 +295,7 @@ def japanese_cleaners(text):
 
 
 def japanese_cleaners2(text):
-  '''Pipeline for notating accent in Japanese text.'''
-  '''Reference https://r9y9.github.io/ttslearn/latest/notebooks/ch10_Recipe-Tacotron.html'''
-  sentences = re.split(_japanese_marks, text)
-  marks = re.findall(_japanese_marks, text)
-  text = ''
-  for i, sentence in enumerate(sentences):
-    if re.match(_japanese_characters, sentence):
-      if text!='':
-        text+=' '
-      labels = pyopenjtalk.extract_fullcontext(sentence)
-      for n, label in enumerate(labels):
-        phoneme = re.search(r'\-([^\+]*)\+', label).group(1)
-        if phoneme not in ['sil','pau']:
-          text += phoneme.replace('ch','ʧ').replace('sh','ʃ').replace('cl','Q').replace('ts','ʦ')
-        else:
-          continue
-        n_moras = int(re.search(r'/F:(\d+)_', label).group(1))
-        a1 = int(re.search(r"/A:(\-?[0-9]+)\+", label).group(1))
-        a2 = int(re.search(r"\+(\d+)\+", label).group(1))
-        a3 = int(re.search(r"\+(\d+)/", label).group(1))
-        if re.search(r'\-([^\+]*)\+', labels[n + 1]).group(1) in ['sil','pau']:
-          a2_next=-1
-        else:
-          a2_next = int(re.search(r"\+(\d+)\+", labels[n + 1]).group(1))
-        # Accent phrase boundary
-        if a3 == 1 and a2_next == 1:
-          text += ' '
-        # Falling
-        elif a1 == 0 and a2_next == a2 + 1 and a2 != n_moras:
-          text += '↓'
-        # Rising
-        elif a2 == 1 and a2_next == 2:
-          text += '↑'
-    if i<len(marks):
-      text += unidecode(marks[i]).replace(' ','')
-  if re.match('[A-Za-z]',text[-1]):
-    text += '.'
-  return text
+  return japanese_cleaners(text).replace('ts','ʦ')
 
 
 def korean_cleaners(text):

train.py

@@ -1,3 +1,4 @@
+from ast import Import
 import os
 import json
 import argparse
@@ -36,7 +37,7 @@
   kl_loss
 )
 from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
-from text.symbols import symbols
+from text import initialize
 
 
 torch.backends.cudnn.benchmark = True
@@ -52,6 +53,7 @@ def main():
   os.environ['MASTER_PORT'] = '80000'
 
   hps = utils.get_hparams()
+  initialize(hps.symbols)
   mp.spawn(run, nprocs=n_gpus, args=(n_gpus, hps,))
 
 
@@ -86,7 +88,7 @@ def run(rank, n_gpus, hps):
         drop_last=False, collate_fn=collate_fn)
 
   net_g = SynthesizerTrn(
-      len(symbols),
+      hps.symbols,
       hps.data.filter_length // 2 + 1,
       hps.train.segment_size // hps.data.hop_length,
       **hps.model).cuda(rank)

train_ms.py

@@ -1,10 +1,5 @@
 import os
-import json
-import argparse
-import itertools
-import math
 import torch
-from torch import nn, optim
 from torch.nn import functional as F
 from torch.utils.data import DataLoader
 from torch.utils.tensorboard import SummaryWriter
@@ -13,7 +8,6 @@
 from torch.nn.parallel import DistributedDataParallel as DDP
 from torch.cuda.amp import autocast, GradScaler
 
-import librosa
 import logging
 
 logging.getLogger('numba').setLevel(logging.WARNING)
@@ -36,7 +30,7 @@
   kl_loss
 )
 from mel_processing import mel_spectrogram_torch, spec_to_mel_torch
-from text.symbols import symbols
+from text import initialize
 
 
 torch.backends.cudnn.benchmark = True
@@ -52,6 +46,7 @@ def main():
   os.environ['MASTER_PORT'] = '80000'
 
   hps = utils.get_hparams()
+  initialize(hps.symbols)
   mp.spawn(run, nprocs=n_gpus, args=(n_gpus, hps,))
 
 
@@ -86,7 +81,7 @@ def run(rank, n_gpus, hps):
         drop_last=False, collate_fn=collate_fn)
 
   net_g = SynthesizerTrn(
-      len(symbols),
+      hps.symbols,
       hps.data.filter_length // 2 + 1,
       hps.train.segment_size // hps.data.hop_length,
       n_speakers=hps.data.n_speakers,