Update inference.py

inference.py

@@ -38,7 +38,7 @@ def get_text(wav,sr,transform=1.0):
         wav16=wav
     
     source = {"source":np.expand_dims(np.expand_dims(wav16,0),0)}
-    hubertsession = onnxruntime.InferenceSession("hubert.onnx")#,providers=['CUDAExecutionProvider'])
+    hubertsession = onnxruntime.InferenceSession("hubert.onnx",providers=['CUDAExecutionProvider'])
     units = np.array(hubertsession.run(['embed'], source)[0])
     f0=_calculate_f0(wav,units.shape[1],sr,
             f0min=librosa.note_to_hz('C2'),
@@ -71,7 +71,7 @@ def infer(f,r,speaker,key,reqf0=False):
     x_lengths = [np.size(x,1)]
     sid = [speaker]
     ort_inputs = {'x':x,'x_lengths':x_lengths,'sid':sid,"noise_scale":[0.667],"length_scale":[1.0],"noise_scale_w":[0.8]} 
-    infersession = onnxruntime.InferenceSession("onnxmodel334.onnx")#,providers=['CUDAExecutionProvider'])
+    infersession = onnxruntime.InferenceSession("onnxmodel334.onnx",providers=['CUDAExecutionProvider'])
     ort_output = infersession.run(['audio'], ort_inputs)
     #sf.write(o,ort_output[0][0][0],22050,'PCM_16',format='wav')
     genf0=np.array([])