update space

app.py

@@ -12,10 +12,11 @@ os.environ['FORCE_NO_DOWNSAMPLE'] = '1'
 os.environ['LOAD_VISION_EARLY'] = '1'
 os.environ['SKIP_LOAD_VIT'] = '1'
 
-os.environ["CUDA_LAUNCH_BLOCKING"]='1'
+# os.environ["CUDA_LAUNCH_BLOCKING"]='1'
 
 import gradio as gr
 import torch
+print(torch.cuda.is_available())
 import re
 from decord import VideoReader, cpu
 from PIL import Image

ola/CosyVoice_main/cosyvoice/cli/model.py

@@ -57,7 +57,8 @@ class CosyVoiceModel:
         # rtf and decoding related
         self.stream_scale_factor = 1
         assert self.stream_scale_factor >= 1, 'stream_scale_factor should be greater than 1, change it according to your actual rtf'
-        self.llm_context = torch.cuda.stream(torch.cuda.Stream(self.device))
+        # self.llm_context = torch.cuda.stream(torch.cuda.Stream(self.device)) if torch.cuda.is_available() else nullcontext()
+        self.llm_context = nullcontext()
         self.lock = threading.Lock()
         # dict used to store session related variable
         self.tts_speech_token_dict = {}