From 5571f76b34f862e74e97e4e6f38645a3f31f26b2 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Stardust=C2=B7=E5=87=8F?= <2225664821@qq.com>
Date: Thu, 3 Aug 2023 18:20:06 +0800
Subject: [PATCH] Update spec_gen.py

---
 spec_gen.py | 27 +++++++++------------------
 1 file changed, 9 insertions(+), 18 deletions(-)

diff --git a/spec_gen.py b/spec_gen.py
index 796cca9..3aee5e3 100644
--- a/spec_gen.py
+++ b/spec_gen.py
@@ -5,28 +5,14 @@ import commons
 import utils
 from data_utils import TextAudioSpeakerLoader, TextAudioSpeakerCollate
 from tqdm import tqdm
+from multiprocessing import Pool
 
 from text import cleaned_text_to_sequence, get_bert
 
 config_path = 'configs/config.json'
 hps = utils.get_hparams_from_file(config_path)
 
-# train_dataset = TextAudioSpeakerLoader(hps.data.training_files, hps.data)
-# eval_dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)
-#
-# collate_fn = TextAudioSpeakerCollate()
-# train_loader = DataLoader(train_dataset, num_workers=12, shuffle=False,
-#                          batch_size=32, pin_memory=True,
-#                          drop_last=False, collate_fn=collate_fn)
-# eval_loader = DataLoader(eval_dataset, num_workers=12, shuffle=False,
-#                          batch_size=32, pin_memory=True,
-#                          drop_last=False, collate_fn=collate_fn)
-# for _ in tqdm(train_loader):
-#     pass
-# for _ in tqdm(eval_loader):
-#     pass
-
-for line in tqdm( open(hps.data.training_files).readlines()):
+def process_line(line):
     _id, spk, language_str, text, phones, tone, word2ph = line.strip().split("|")
     phone = phones.split(" ")
     tone = [int(i) for i in tone.split(" ")]
@@ -35,7 +21,6 @@ for line in tqdm( open(hps.data.training_files).readlines()):
     w2pho = [i for i in word2ph]
     word2ph = [i for i in word2ph]
     phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str)
-    pold2 = phone
 
     if hps.data.add_blank:
         phone = commons.intersperse(phone, 0)
@@ -44,7 +29,7 @@ for line in tqdm( open(hps.data.training_files).readlines()):
         for i in range(len(word2ph)):
             word2ph[i] = word2ph[i] * 2
         word2ph[0] += 1
-    wav_path = f'dataset/{spk}/{_id}.wav'
+    wav_path = f'{_id}'
 
     bert_path = wav_path.replace(".wav", ".bert.pt")
     try:
@@ -55,3 +40,9 @@ for line in tqdm( open(hps.data.training_files).readlines()):
         assert bert.shape[-1] == len(phone)
         torch.save(bert, bert_path)
 
+with open(hps.data.training_files) as f:
+    lines = f.readlines()
+
+with Pool() as pool:
+    for _ in tqdm(pool.imap_unordered(process_line, lines)):
+        pass