(wip)add gpu tags

app.py

@@ -629,6 +629,7 @@ def generate_tts():
         model_ids = []
 
         # Function to process a single model (generate directly to TEMP_AUDIO_DIR, not cache subdir)
+        @spaces.GPU
         def process_model_on_the_fly(model):
              # 传递 reference_audio_path 给 predict_tts
              temp_audio_path = predict_tts(text, model.id, reference_audio_path=reference_audio_path)

cosyvoice/cli/cosyvoice.py

@@ -16,7 +16,7 @@ import spaces
 import time
 from tqdm import tqdm
 from hyperpyyaml import load_hyperpyyaml
-from modelscope import snapshot_download
+from huggingface_hub import snapshot_download
 import torch
 from cosyvoice.cli.cosyvoice import CosyVoiceFrontEnd
 from cosyvoice.cli.cosyvoice import CosyVoiceModel, CosyVoice2Model

tts.py

@@ -236,10 +236,10 @@ def predict_spark_tts(text, reference_audio_path=None):
 def predict_cosyvoice_tts(text, reference_audio_path=None):
     import tempfile
     import soundfile as sf
-    from modelscope import snapshot_download
+    from huggingface_hub import snapshot_download
     model_dir = os.path.join(os.path.dirname(__file__), "CosyVoice2-0.5B", "pretrained_models", "CosyVoice2-0.5B")
     if not os.path.exists(model_dir) or not os.listdir(model_dir):
-        snapshot_download('iic/CosyVoice2-0.5B', local_dir=model_dir)
+        snapshot_download('FunAudioLLM/CosyVoice2-0.5B', local_dir=model_dir)
     sys.path.append(os.path.join(os.path.dirname(__file__), "CosyVoice2-0.5B"))
     from cosyvoice.cli.cosyvoice import CosyVoice2
     from cosyvoice.utils.file_utils import load_wav