feat: install

GPT_SoVITS/inference_webui.py

@@ -7,6 +7,7 @@
 全部按日文识别
 """
 
+import spaces
 from module.mel_processing import mel_spectrogram_torch, spectrogram_torch
 from text import chinese
 from sv import SV
@@ -34,12 +35,30 @@ import re
 import sys
 import traceback
 import warnings
-import spaces
-
 import torch
 import torchaudio
 from text.LangSegmenter import LangSegmenter
 
+import zipfile
+from huggingface_hub import hf_hub_download
+import nltk
+nltk.download(['averaged_perceptron_tagger',
+              'averaged_perceptron_tagger_eng', 'cmudict'])
+
+zip_targets = {
+    "pretrained_models.zip": "/home/user/app",
+    "G2PWModel.zip": "/home/user/app/text"
+}
+repo_id = "XXXXRT/GPT-SoVITS-Pretrained"
+
+for filename, target_dir in zip_targets.items():
+    zip_path = hf_hub_download(repo_id=repo_id, filename=filename)
+    os.makedirs(target_dir, exist_ok=True)
+    with zipfile.ZipFile(zip_path, "r") as zip_ref:
+        zip_ref.extractall(target_dir)
+    print(f"{filename} 已解压到 {target_dir}")
+
+
 # 保存原始构造器
 original_storage_new = torch.UntypedStorage.__new__
 
@@ -76,7 +95,7 @@ logging.getLogger("charset_normalizer").setLevel(logging.ERROR)
 logging.getLogger("torchaudio._extension").setLevel(logging.ERROR)
 logging.getLogger("multipart.multipart").setLevel(logging.ERROR)
 warnings.simplefilter(action="ignore", category=FutureWarning)
-os.system("bash install.sh")
+# os.system("bash install.sh")
 version = model_version = os.environ.get("version", "v2")
 
 
@@ -770,6 +789,8 @@ def audio_sr(audio, sr):
 cache = {}
 
 
+@torch.inference_mode()
+@spaces.GPU
 def get_tts_wav(
     ref_wav_path,
     prompt_text,

README.md

@@ -8,6 +8,9 @@ sdk_version: 5.33.0
 app_file: GPT_SoVITS/inference_webui.py
 pinned: false
 license: apache-2.0
+preload_from_hub:
+    - XXXXRT/GPT-SoVITS-Pretrained pretrained_models.zip
+    - XXXXRT/GPT-SoVITS-Pretrained G2PWModel.zip
 ---
 
 Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference