Update app.py

app.py

@@ -1,13 +1,13 @@
 import gradio as gr
-from transformers import AutoTokenizer, AutoModel
+from transformers import AutoTokenizer, AutoModelForCausalLM
 
 tokenizer = AutoTokenizer.from_pretrained("HIT-TMG/bge-m3_RAG-conversational-IR")
-model = AutoModel.from_pretrained("HIT-TMG/bge-m3_RAG-conversational-IR")
+model = AutoModelForCausalLM.from_pretrained("HIT-TMG/bge-m3_RAG-conversational-IR")
 
 def predict(input_text):
     inputs = tokenizer(input_text, return_tensors="pt")
-    outputs = model(**inputs)
-    return outputs.last_hidden_state[0][0].tolist()
+    outputs = model.generate(**inputs, max_length=50)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
 iface = gr.Interface(fn=predict, inputs="text", outputs="text")
 iface.launch()