add

app.py

@@ -51,10 +51,10 @@ pipe = pipeline(
   "automatic-speech-recognition",
   model="openai/whisper-tiny.en",
   chunk_length_s=30,
-  device=device,
+  device='cpu',
 )
        
-@spaces.GPU()
+# @spaces.GPU()
 def run_pipeline(audio):
     return pipe(audio, batch_size=8)["text"]
 
@@ -333,12 +333,10 @@ class BaseTrainer(object):
         return result
 
 
-@spaces.GPU(duration=100)
+@spaces.GPU(duration=140)
 def _warp(args,model, batch_data,joints,joint_mask_upper,joint_mask_hands,joint_mask_lower,use_trans,mean_upper,mean_hands,mean_lower,std_upper,std_hands,std_lower,trans_mean,trans_std):
     diffusion = create_gaussian_diffusion(use_ddim=args.use_ddim)
     args,model,vq_model_upper,vq_model_hands,vq_model_lower,mean_upper,mean_hands,mean_lower,std_upper,std_hands,std_lower,trans_mean,trans_std,vqvae_latent_scale=_warp_create_cuda_model(args,model,mean_upper,mean_hands,mean_lower,std_upper,std_hands,std_lower,trans_mean,trans_std)
-    
-    
     loaded_data = _warp_load_data(
         batch_data,joints,joint_mask_upper,joint_mask_hands,joint_mask_lower,args,use_trans,mean_upper,mean_hands,mean_lower,std_upper,std_hands,std_lower,trans_mean,trans_std,vq_model_upper,vq_model_hands,vq_model_lower        
     )