update

TTS_infer_pack/TTS.py

@@ -149,45 +149,45 @@ class NO_PROMPT_ERROR(Exception):
 # configs/tts_infer.yaml
 """
 custom:
-  bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
-  cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
+  bert_base_path: pretrained_models/chinese-roberta-wwm-ext-large
+  cnhuhbert_base_path: pretrained_models/chinese-hubert-base
   device: cpu
   is_half: false
-  t2s_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
-  vits_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
+  t2s_weights_path: pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
+  vits_weights_path: pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
   version: v2
 v1:
-  bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
-  cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
+  bert_base_path: pretrained_models/chinese-roberta-wwm-ext-large
+  cnhuhbert_base_path: pretrained_models/chinese-hubert-base
   device: cpu
   is_half: false
-  t2s_weights_path: GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt
-  vits_weights_path: GPT_SoVITS/pretrained_models/s2G488k.pth
+  t2s_weights_path: pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt
+  vits_weights_path: pretrained_models/s2G488k.pth
   version: v1
 v2:
-  bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
-  cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
+  bert_base_path: pretrained_models/chinese-roberta-wwm-ext-large
+  cnhuhbert_base_path: pretrained_models/chinese-hubert-base
   device: cpu
   is_half: false
-  t2s_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
-  vits_weights_path: GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
+  t2s_weights_path: pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
+  vits_weights_path: pretrained_models/gsv-v2final-pretrained/s2G2333k.pth
   version: v2
 v3:
-  bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
-  cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
+  bert_base_path: pretrained_models/chinese-roberta-wwm-ext-large
+  cnhuhbert_base_path: pretrained_models/chinese-hubert-base
   device: cpu
   is_half: false
-  t2s_weights_path: GPT_SoVITS/pretrained_models/s1v3.ckpt
-  vits_weights_path: GPT_SoVITS/pretrained_models/s2Gv3.pth
+  t2s_weights_path: pretrained_models/s1v3.ckpt
+  vits_weights_path: pretrained_models/s2Gv3.pth
   version: v3
 v4:
-  bert_base_path: GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large
-  cnhuhbert_base_path: GPT_SoVITS/pretrained_models/chinese-hubert-base
+  bert_base_path: pretrained_models/chinese-roberta-wwm-ext-large
+  cnhuhbert_base_path: pretrained_models/chinese-hubert-base
   device: cpu
   is_half: false
-  t2s_weights_path: GPT_SoVITS/pretrained_models/s1v3.ckpt
+  t2s_weights_path: pretrained_models/s1v3.ckpt
   version: v4
-  vits_weights_path: GPT_SoVITS/pretrained_models/gsv-v4-pretrained/s2Gv4.pth
+  vits_weights_path: pretrained_models/gsv-v4-pretrained/s2Gv4.pth
 """
 
 
@@ -220,55 +220,55 @@ class TTS_Config:
             "device": "cpu",
             "is_half": False,
             "version": "v1",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/s2G488k.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt",
+            "vits_weights_path": "pretrained_models/s2G488k.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
         "v2": {
             "device": "cpu",
             "is_half": False,
             "version": "v2",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt",
+            "vits_weights_path": "pretrained_models/gsv-v2final-pretrained/s2G2333k.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
         "v3": {
             "device": "cpu",
             "is_half": False,
             "version": "v3",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/s2Gv3.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/s1v3.ckpt",
+            "vits_weights_path": "pretrained_models/s2Gv3.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
         "v4": {
             "device": "cpu",
             "is_half": False,
             "version": "v4",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/gsv-v4-pretrained/s2Gv4.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/s1v3.ckpt",
+            "vits_weights_path": "pretrained_models/gsv-v4-pretrained/s2Gv4.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
         "v2Pro": {
             "device": "cpu",
             "is_half": False,
             "version": "v2Pro",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/v2Pro/s2Gv2Pro.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/s1v3.ckpt",
+            "vits_weights_path": "pretrained_models/v2Pro/s2Gv2Pro.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
         "v2ProPlus": {
             "device": "cpu",
             "is_half": False,
             "version": "v2ProPlus",
-            "t2s_weights_path": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
-            "vits_weights_path": "GPT_SoVITS/pretrained_models/v2Pro/s2Gv2ProPlus.pth",
-            "cnhuhbert_base_path": "GPT_SoVITS/pretrained_models/chinese-hubert-base",
-            "bert_base_path": "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+            "t2s_weights_path": "pretrained_models/s1v3.ckpt",
+            "vits_weights_path": "pretrained_models/v2Pro/s2Gv2ProPlus.pth",
+            "cnhuhbert_base_path": "pretrained_models/chinese-hubert-base",
+            "bert_base_path": "pretrained_models/chinese-roberta-wwm-ext-large",
         },
     }
     configs: dict = None
@@ -289,7 +289,7 @@ class TTS_Config:
 
     def __init__(self, configs: Union[dict, str] = None):
         # 设置默认配置文件路径
-        configs_base_path: str = "GPT_SoVITS/configs/"
+        configs_base_path: str = "configs/"
         os.makedirs(configs_base_path, exist_ok=True)
         self.configs_path: str = os.path.join(configs_base_path, "tts_infer.yaml")
 
@@ -602,7 +602,7 @@ class TTS:
                 self.empty_cache()
 
             self.vocoder = BigVGAN.from_pretrained(
-                "%s/GPT_SoVITS/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
+                "%s/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
                 use_cuda_kernel=False,
             )  # if True, RuntimeError: Ninja is required to load C++ extensions
             # remove weight norm in the model and set to eval mode
@@ -635,7 +635,7 @@ class TTS:
             )
             self.vocoder.remove_weight_norm()
             state_dict_g = torch.load(
-                "%s/GPT_SoVITS/pretrained_models/gsv-v4-pretrained/vocoder.pth" % (now_dir,),
+                "%s/pretrained_models/gsv-v4-pretrained/vocoder.pth" % (now_dir,),
                 map_location="cpu",
                 weights_only=False,
             )

TTS_infer_pack/text_segmentation_method.py

@@ -158,7 +158,7 @@ def cut4(inp):
 
 
 # 按标点符号切
-# contributed by https://github.com/AI-Hobbyist/GPT-SoVITS/blob/main/GPT_SoVITS/inference_webui.py
+# contributed by https://github.com/AI-Hobbyist/GPT-SoVITS/blob/main/inference_webui.py
 @register_method("cut5")
 def cut5(inp):
     inp = inp.strip("\n")

download.py

@@ -6,8 +6,8 @@ sys.path.insert(0, now_dir)
 from text.g2pw import G2PWPinyin
 
 g2pw = G2PWPinyin(
-    model_dir="GPT_SoVITS/text/G2PWModel",
-    model_source="GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large",
+    model_dir="text/G2PWModel",
+    model_source="pretrained_models/chinese-roberta-wwm-ext-large",
     v_to_u=False,
     neutral_tone_with_five=True,
 )

export_torch_script.py

@@ -561,8 +561,8 @@ class T2SModel(nn.Module):
         return y[:, -idx:].unsqueeze(0)
 
 
-bert_path = os.environ.get("bert_path", "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large")
-cnhubert_base_path = "GPT_SoVITS/pretrained_models/chinese-hubert-base"
+bert_path = os.environ.get("bert_path", "pretrained_models/chinese-roberta-wwm-ext-large")
+cnhubert_base_path = "pretrained_models/chinese-hubert-base"
 cnhubert.cnhubert_base_path = cnhubert_base_path
 
 

export_torch_script_v3v4.py

@@ -505,7 +505,7 @@ def init_bigvgan():
     from BigVGAN import bigvgan
 
     bigvgan_model = bigvgan.BigVGAN.from_pretrained(
-        "%s/GPT_SoVITS/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
+        "%s/pretrained_models/models--nvidia--bigvgan_v2_24khz_100band_256x" % (now_dir,),
         use_cuda_kernel=False,
     )  # if True, RuntimeError: Ninja is required to load C++ extensions
     # remove weight norm in the model and set to eval mode
@@ -533,7 +533,7 @@ def init_hifigan():
     hifigan_model.eval()
     hifigan_model.remove_weight_norm()
     state_dict_g = torch.load(
-        "%s/GPT_SoVITS/pretrained_models/gsv-v4-pretrained/vocoder.pth" % (now_dir,), map_location="cpu"
+        "%s/pretrained_models/gsv-v4-pretrained/vocoder.pth" % (now_dir,), map_location="cpu"
     )
     print("loading vocoder", hifigan_model.load_state_dict(state_dict_g))
     if is_half == True:
@@ -584,7 +584,7 @@ v3v4set = {"v3", "v4"}
 
 
 def get_sovits_weights(sovits_path):
-    path_sovits_v3 = "GPT_SoVITS/pretrained_models/s2Gv3.pth"
+    path_sovits_v3 = "pretrained_models/s2Gv3.pth"
     is_exist_s2gv3 = os.path.exists(path_sovits_v3)
 
     version, model_version, if_lora_v3 = get_sovits_version_from_path_fast(sovits_path)
@@ -707,13 +707,13 @@ def export_cfm(
 
 def export_1(ref_wav_path, ref_wav_text, version="v3"):
     if version == "v3":
-        sovits = get_sovits_weights("GPT_SoVITS/pretrained_models/s2Gv3.pth")
+        sovits = get_sovits_weights("pretrained_models/s2Gv3.pth")
         init_bigvgan()
     else:
-        sovits = get_sovits_weights("GPT_SoVITS/pretrained_models/gsv-v4-pretrained/s2Gv4.pth")
+        sovits = get_sovits_weights("pretrained_models/gsv-v4-pretrained/s2Gv4.pth")
         init_hifigan()
 
-    dict_s1 = torch.load("GPT_SoVITS/pretrained_models/s1v3.ckpt")
+    dict_s1 = torch.load("pretrained_models/s1v3.ckpt")
     raw_t2s = get_raw_t2s_model(dict_s1).to(device)
     print("#### get_raw_t2s_model ####")
     print(raw_t2s.config)
@@ -1124,10 +1124,10 @@ import time
 
 def export_2(version="v3"):
     if version == "v3":
-        sovits = get_sovits_weights("GPT_SoVITS/pretrained_models/s2Gv3.pth")
+        sovits = get_sovits_weights("pretrained_models/s2Gv3.pth")
         # init_bigvgan()
     else:
-        sovits = get_sovits_weights("GPT_SoVITS/pretrained_models/gsv-v4-pretrained/s2Gv4.pth")
+        sovits = get_sovits_weights("pretrained_models/gsv-v4-pretrained/s2Gv4.pth")
         # init_hifigan()
 
     # cfm = ExportCFM(sovits.cfm)
@@ -1142,9 +1142,9 @@ def export_2(version="v3"):
 
     logger.info("cfm ok")
 
-    dict_s1 = torch.load("GPT_SoVITS/pretrained_models/s1v3.ckpt")
+    dict_s1 = torch.load("pretrained_models/s1v3.ckpt")
     # v2 的 gpt 也可以用
-    # dict_s1 = torch.load("GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt")
+    # dict_s1 = torch.load("pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt")
     raw_t2s = get_raw_t2s_model(dict_s1).to(device)
     print("#### get_raw_t2s_model ####")
     print(raw_t2s.config)

inference_webui.py

@@ -1063,7 +1063,7 @@ def cut4(inp):
     return "\n".join(opts)
 
 
-# contributed by https://github.com/AI-Hobbyist/GPT-SoVITS/blob/main/GPT_SoVITS/inference_webui.py
+# contributed by https://github.com/AI-Hobbyist/GPT-SoVITS/blob/main/inference_webui.py
 def cut5(inp):
     inp = inp.strip("\n")
     punds = {",", ".", ";", "?", "!", "、", "，", "。", "？", "！", ";", "：", "…"}

inference_webui_fast.py

@@ -109,7 +109,7 @@ path_sovits_v4 = pretrained_sovits_name["v4"]
 is_exist_s2gv3 = os.path.exists(path_sovits_v3)
 is_exist_s2gv4 = os.path.exists(path_sovits_v4)
 
-tts_config = TTS_Config("GPT_SoVITS/configs/tts_infer.yaml")
+tts_config = TTS_Config("configs/tts_infer.yaml")
 tts_config.device = device
 tts_config.is_half = is_half
 tts_config.version = version

onnx_export.py

@@ -5,7 +5,7 @@ from feature_extractor import cnhubert
 from module.models_onnx import SynthesizerTrn, symbols_v1, symbols_v2
 from torch import nn
 
-cnhubert_base_path = "GPT_SoVITS/pretrained_models/chinese-hubert-base"
+cnhubert_base_path = "pretrained_models/chinese-hubert-base"
 cnhubert.cnhubert_base_path = cnhubert_base_path
 ssl_model = cnhubert.get_model()
 import json

prepare_datasets/2-get-sv.py

@@ -22,7 +22,7 @@ import torchaudio
 
 now_dir = os.getcwd()
 sys.path.append(now_dir)
-sys.path.append(f"{now_dir}/GPT_SoVITS/eres2net")
+sys.path.append(f"{now_dir}/eres2net")
 from tools.my_utils import clean_path
 from time import time as ttime
 import shutil

sv.py

@@ -2,8 +2,8 @@ import sys
 import os
 import torch
 
-sys.path.append(f"{os.getcwd()}/GPT_SoVITS/eres2net")
-sv_path = "GPT_SoVITS/pretrained_models/sv/pretrained_eres2netv2w24s4ep4.ckpt"
+sys.path.append(f"{os.getcwd()}/eres2net")
+sv_path = "pretrained_models/sv/pretrained_eres2netv2w24s4ep4.ckpt"
 from ERes2NetV2 import ERes2NetV2
 import kaldi as Kaldi
 

text/chinese2.py

@@ -32,8 +32,8 @@ if is_g2pw:
 
     parent_directory = os.path.dirname(current_file_path)
     g2pw = G2PWPinyin(
-        model_dir="GPT_SoVITS/text/G2PWModel",
-        model_source=os.environ.get("bert_path", "GPT_SoVITS/pretrained_models/chinese-roberta-wwm-ext-large"),
+        model_dir="text/G2PWModel",
+        model_source=os.environ.get("bert_path", "pretrained_models/chinese-roberta-wwm-ext-large"),
         v_to_u=False,
         neutral_tone_with_five=True,
     )