Update app.py

app.py

@@ -65,47 +65,7 @@ with gr.Blocks() as demo:
                 yield '(finished initialization, responding...)', state
                 if mode != 'Conversation':
                     yield 'waiting video input...', state
-                query = message
-                if video_path:
-                    message = {
-                        "role": "user",
-                        "content": [
-                            {"type": "video", "video": video_path},
-                            {"type": "text", "text": query},
-                        ],
-                    }
-                    
-                else:
-                    message = {
-                        "role": "user",
-                        "content": [
-                            {"type": "text", "text": query},
-                        ],
-                    }
-                image_inputs, video_inputs = process_vision_info([message])
-                texts = infer.processor.apply_chat_template([message], tokenize=False, add_generation_prompt=True, return_tensors='pt')
-                past_ids = state.get('past_ids', None)
-                if past_ids is not None:
-                    texts = '<|im_end|>\n' + texts[infer.system_prompt_offset:]
-                inputs = infer.processor(
-                    text=texts,
-                    images=image_inputs,
-                    videos=video_inputs,
-                    return_tensors="pt",
-                )
-                inputs.to(infer.model.device)
-                if past_ids is not None:
-                    inputs['input_ids'] = torch.cat([past_ids, inputs.input_ids], dim=1) 
-                outputs = infer.model.generate(
-                    **inputs, past_key_values=state.get('past_key_values', None), 
-                    return_dict_in_generate=True, do_sample=False, 
-                    repetition_penalty=1.05,
-                    max_new_tokens=512,
-                )
-                state['past_key_values'] = outputs.past_key_values
-                state['past_ids'] = outputs.sequences[:, :-1]
-                response = infer.processor.decode(outputs.sequences[0, inputs.input_ids.size(1):], skip_special_tokens=True)
-                print(response)
+                response, _ = infer.video_qa(query=message, state=state)
                 yield response, {}                
                 
             def gr_chatinterface_chatbot_clear_fn():