Format code

data_utils.py

@@ -208,7 +208,13 @@ def __call__(self, batch):
             torch.LongTensor([x[1].size(1) for x in batch]), dim=0, descending=True
         )
 
-        max_text_len = max([batch[ids_sorted_decreasing[i]][7].size(1) for i in range(len(ids_sorted_decreasing))] + [len(x[0]) for x in batch])
+        max_text_len = max(
+            [
+                batch[ids_sorted_decreasing[i]][7].size(1)
+                for i in range(len(ids_sorted_decreasing))
+            ]
+            + [len(x[0]) for x in batch]
+        )
         max_spec_len = max([x[1].size(1) for x in batch])
         max_wav_len = max([x[2].size(1) for x in batch])
 

text/__init__.py

@@ -11,7 +11,7 @@ def cleaned_text_to_sequence(cleaned_text, tones, language):
     Returns:
       List of integers corresponding to the symbols in the text
     """
-    phones = [] # _symbol_to_id[symbol] for symbol in cleaned_text
+    phones = []  # _symbol_to_id[symbol] for symbol in cleaned_text
     for symbol in cleaned_text:
         try:
             phones.append(_symbol_to_id[symbol])

text/japanese.py

@@ -11,7 +11,8 @@
 import pyopenjtalk
 
 from num2words import num2words
-BERT = './bert/bert-large-japanese-v2'
+
+BERT = "./bert/bert-large-japanese-v2"
 _CONVRULES = [
     # Conversion of 2 letters
     "アァ/ a a",
@@ -385,6 +386,7 @@ def replace_punctuation(text):
         replaced_text = replaced_text.replace(x, rep_map[x])
     return replaced_text
 
+
 def text_normalize(text):
     res = unicodedata.normalize("NFKC", text)
     res = japanese_convert_numbers_to_words(res)
@@ -393,27 +395,29 @@ def text_normalize(text):
     return res
 
 
-
 tokenizer = AutoTokenizer.from_pretrained(BERT)
 
+
 def g2p(norm_text):
     tokenized = tokenizer.tokenize(norm_text)
     st = [x.replace("#", "") for x in tokenized]
     word2ph = []
-    phs = pyopenjtalk.g2p(norm_text).split(" ")  # Directly use the entire norm_text sequence.
+    phs = pyopenjtalk.g2p(norm_text).split(
+        " "
+    )  # Directly use the entire norm_text sequence.
     for sub in st:  # the following code is only for calculating word2ph
         wph = 0
         for x in sub:
             sys.stdout.flush()
-            if x not in ['?', '.', '!', '…', ',']:  # This will throw warnings.
+            if x not in ["?", ".", "!", "…", ","]:  # This will throw warnings.
                 phonemes = pyopenjtalk.g2p(x)
             else:
-                phonemes = 'pau'
+                phonemes = "pau"
             # for x in range(repeat):
-            wph += len(phonemes.split(' '))
+            wph += len(phonemes.split(" "))
             # print(f'{x}-->:{phones}')
         word2ph.append(wph)
-    phonemes = ['_'] + phs + ['_']
+    phonemes = ["_"] + phs + ["_"]
     tones = [0 for i in phonemes]
     word2ph = [1] + word2ph + [1]
     return phonemes, tones, word2ph

text/japanese_bert.py

@@ -2,7 +2,7 @@
 from transformers import AutoTokenizer, AutoModelForMaskedLM
 import sys
 
-BERT = './bert/bert-large-japanese-v2'
+BERT = "./bert/bert-large-japanese-v2"
 
 tokenizer = AutoTokenizer.from_pretrained(BERT)
 # bert-large model has 25 hidden layers.You can decide which layer to use by setting this variable to a specific value
@@ -19,15 +19,13 @@ def get_bert_feature(text, word2ph, device=None):
         device = "mps"
     if not device:
         device = "cuda"
-    model = AutoModelForMaskedLM.from_pretrained(BERT).to(
-        device
-    )
+    model = AutoModelForMaskedLM.from_pretrained(BERT).to(device)
     with torch.no_grad():
         inputs = tokenizer(text, return_tensors="pt")
         for i in inputs:
             inputs[i] = inputs[i].to(device)
         res = model(**inputs, output_hidden_states=True)
-        res = res['hidden_states'][BERT_LAYER]
+        res = res["hidden_states"][BERT_LAYER]
     assert inputs["input_ids"].shape[-1] == len(word2ph)
     word2phone = word2ph
     phone_level_feature = []

text/symbols.py

@@ -118,7 +118,7 @@
     "z",
     "zy",
 ]
-for x in range(ord('a'), ord('z')+1):
+for x in range(ord("a"), ord("z") + 1):
     ja_symbols.append(chr(x).upper())
 num_ja_tones = 1