Update demo/infer.py

demo/infer.py

@@ -216,4 +216,5 @@ class LiveCCDemoInfer:
         state['past_ids'] = outputs.sequences[:, :-1]
         response = self.processor.decode(outputs.sequences[0, inputs.input_ids.size(1):], skip_special_tokens=True)
         print(response)
+        state.pop('past_key_values')
         return response, state