提交代码更改

utils/model_inference.py

@@ -38,6 +38,7 @@ model = MultimodalClassifier().to(device)
 
 # 加载 SafeTensors 权重
 state_dict = load_file(model_path)
+print("state_dict:", state_dict)
 model.load_state_dict(state_dict)  
 model.eval()  # 设置为评估模式
 
@@ -58,7 +59,7 @@ def predict_emotion(text, audio):
     audio_inputs = preprocess_audio(audio)
 
     with torch.no_grad():
-        output = model(text_inputs, audio_inputs)  # (1, 7) logits
+        output = model(audio_inputs,text_inputs)  # (1, 7) logits
         probabilities = F.softmax(output, dim=1).squeeze().tolist()  # 归一化为概率
 
     return {labels[i]: f"{probabilities[i]*100:.2f}%" for i in range(len(labels))}