update

app.py

@@ -24,12 +24,17 @@ print(model_id,device,dtype)
 histories = []
 #model = None
 
+model = AutoModelForCausalLM.from_pretrained(
+        model_id, token=huggingface_token ,torch_dtype=dtype,device_map=device
+    )
+text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer,torch_dtype=dtype,device_map=device) #pipeline has not to(device)
+    
 
 @spaces.GPU(duration=120)
 def generate_text(messages):
-    model = AutoModelForCausalLM.from_pretrained(
-        model_id, token=huggingface_token ,torch_dtype=dtype,device_map=device
-    )
+#    model = AutoModelForCausalLM.from_pretrained(
+#       model_id, token=huggingface_token ,torch_dtype=dtype,device_map=device
+#  )
 
     text_generator = pipeline("text-generation", model=model, tokenizer=tokenizer,torch_dtype=dtype,device_map=device) #pipeline has not to(device)
     result = text_generator(messages, max_new_tokens=256, do_sample=True, temperature=0.7)
@@ -48,7 +53,7 @@ def generate_text(messages):
 
 
 def call_generate_text(message, history):
-    history.append({"role": "assistant", "content": message})
+    history.append({"role": "user", "content": message})
     print(message)
     print(history)
    
@@ -63,7 +68,7 @@ def call_generate_text(message, history):
        
     return ""
 
-demo = gr.ChatInterface(call_generate_text)
+demo = gr.ChatInterface(call_generate_text,type="messages")
 
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(share=True)