Update app.py

app.py

@@ -3,14 +3,29 @@ import torch
 import torchaudio
 from torchaudio.transforms import Resample
 
+# 定义一个简化的模型类（假设模型是LSTM架构）
+class ASRModel(torch.nn.Module):
+    def __init__(self):
+        super(ASRModel, self).__init__()
+        self.lstm = torch.nn.LSTM(input_size=160, hidden_size=256, num_layers=3, batch_first=True)
+        self.linear = torch.nn.Linear(256, 29)  # 假设29个输出类用于字符
+
+    def forward(self, x):
+        x, _ = self.lstm(x)
+        x = self.linear(x)
+        return x
+
 # 定义模型路径
 model_path = "https://huggingface.co/Tele-AI/TeleSpeech-ASR1.0/resolve/main/large.pt"
 
 # 下载模型文件
 torch.hub.download_url_to_file(model_path, 'large.pt')
 
+# 初始化模型
+model = ASRModel()
+
 # 加载模型参数
-model = torch.load('large.pt', map_location=torch.device('cpu'))
+model.load_state_dict(torch.load('large.pt', map_location=torch.device('cpu')))
 model.eval()
 
 # 定义处理函数
@@ -23,7 +38,7 @@ def transcribe(audio):
     with torch.no_grad():
         logits = model(input_values)
     predicted_ids = torch.argmax(logits, dim=-1)
-    transcription = tokenizer.decode(predicted_ids[0])
+    transcription = ''.join([chr(i) for i in predicted_ids[0].tolist()])  # 解码预测到字符
     return transcription
 
 # 创建 Gradio 界面