Update app.py

app.py

@@ -7,19 +7,29 @@ import matplotlib.pyplot as plt
 device="cpu"
 bundle = torchaudio.pipelines.TACOTRON2_WAVERNN_PHONE_LJSPEECH
 processor = bundle.get_text_processor()
-tacotron2 = bundle.get_tacotron2().to(device)
+# tacotron2 = bundle.get_tacotron2().to(device)
+
+tacotron2 = torch.hub.load(
+    "NVIDIA/DeepLearningExamples:torchhub",
+    "nvidia_tacotron2",
+    pretrained=False,
+)
 
 # Load Weights and bias of nepali text
 checkpoint_path = os.path.join(os.getcwd(), 'model_E45.ckpt')
-checkpoint = torch.load(checkpoint_path, map_location=device)
+state_dict = torch.load(checkpoint_path, map_location=device)
+
+tacotron2.load_state_dict(state_dict)
+tacotron2 = tacotron2.to(device)
+tacotron2.eval()
 
-tts_state_dict_stage1 = {key.replace("conv.", ""): value for key, value in checkpoint.items()}
-tts_state_dict_stage2 = {key.replace("linear_layer.", ""): value for key, value in tts_state_dict_stage1.items()}
+# tts_state_dict_stage1 = {key.replace("conv.", ""): value for key, value in checkpoint.items()}
+# tts_state_dict_stage2 = {key.replace("linear_layer.", ""): value for key, value in tts_state_dict_stage1.items()}
 
-# decoder.attention_layer.location_layer.location_weight  --> decoder.attention_layer.location_layer.location_conv.weight
-tts_state_dict_stage3 = {key.replace("location_weight", "location_conv.weight"): value for key, value in tts_state_dict_stage2.items()}
+# # decoder.attention_layer.location_layer.location_weight  --> decoder.attention_layer.location_layer.location_conv.weight
+# tts_state_dict_stage3 = {key.replace("location_weight", "location_conv.weight"): value for key, value in tts_state_dict_stage2.items()}
 
-tacotron2.load_state_dict(tts_state_dict_stage3)
+# tacotron2.load_state_dict(tts_state_dict_stage3)
 
 # Workaround to load model mapped on GPU
 # https://stackoverflow.com/a/61840832