SyntaxError: invalid syntax

[tranvanhoa533]

Describe the issue

Issue: I try to run your given example: Detect the person and frisbee in the image in detection examples. Sometime i meet error: SyntaxError: invalid syntax

Log:

2023-11-14 14:25:00 | INFO | stdout | template_name:  llava_v0
2023-11-14 14:25:00 | INFO | stdout | Messages： [['Human', ('Detect the person and frisbee in the image.\n<image>', <PIL.Image.Image image mode=RGB size=640x428 at 0x7FDD7DFAC5B0>, 'Crop', <PIL.PngImagePlugin.PngImageFile image mode=RGBA size=640x428 at 0x7FDD7DFAC880>)], ['Assistant', None]]
2023-11-14 14:25:00 | INFO | gradio_web_server | model_name: llava_plus_v0_7b, worker_addr: http://localhost:40000
2023-11-14 14:25:00 | INFO | gradio_web_server | ==== request ====
{'model': 'llava_plus_v0_7b', 'prompt': "A chat between a curious human and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the human's questions.###Human: <image>\nDetect the person and frisbee in the image.###Assistant:", 'temperature': 0.2, 'top_p': 0.7, 'max_new_tokens': 512, 'stop': '###', 'images': "List of 1 images: ['e75f292184a71c98df096cba7e880afa']"}
==== request ====
2023-11-14 14:25:05 | ERROR | stderr | Traceback (most recent call last):
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/routes.py", line 437, in run_predict
2023-11-14 14:25:05 | ERROR | stderr |     output = await app.get_blocks().process_api(
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/blocks.py", line 1352, in process_api
2023-11-14 14:25:05 | ERROR | stderr |     result = await self.call_function(
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/blocks.py", line 1093, in call_function
2023-11-14 14:25:05 | ERROR | stderr |     prediction = await utils.async_iteration(iterator)
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/utils.py", line 341, in async_iteration
2023-11-14 14:25:05 | ERROR | stderr |     return await iterator.__anext__()
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/utils.py", line 334, in __anext__
2023-11-14 14:25:05 | ERROR | stderr |     return await anyio.to_thread.run_sync(
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/anyio/to_thread.py", line 33, in run_sync
2023-11-14 14:25:05 | ERROR | stderr |     return await get_asynclib().run_sync_in_worker_thread(
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 877, in run_sync_in_worker_thread
2023-11-14 14:25:05 | ERROR | stderr |     return await future
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/anyio/_backends/_asyncio.py", line 807, in run
2023-11-14 14:25:05 | ERROR | stderr |     result = context.run(func, *args)
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/miniconda3/envs/llava/lib/python3.10/site-packages/gradio/utils.py", line 317, in run_sync_iterator_async
2023-11-14 14:25:05 | ERROR | stderr |     return next(iterator)
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/hoatv2/codes/LLaVA-Plus/llava/serve/gradio_web_server_llava_plus.py", line 471, in http_bot
2023-11-14 14:25:05 | ERROR | stderr |     yield (state, state.to_gradio_chatbot(with_debug_parameter=with_debug_parameter_from_state)) + (disable_btn,) * 6
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/hoatv2/codes/LLaVA-Plus/llava/conversation.py", line 435, in to_gradio_chatbot
2023-11-14 14:25:05 | ERROR | stderr |     ret = self.merge_output(ret, with_debug_parameter=with_debug_parameter)
2023-11-14 14:25:05 | ERROR | stderr |   File "/data/hoatv2/codes/LLaVA-Plus/llava/conversation.py", line 296, in merge_output
2023-11-14 14:25:05 | ERROR | stderr |     action_json = eval(action)
2023-11-14 14:25:05 | ERROR | stderr |   File "<string>", line 2
2023-11-14 14:25:05 | ERROR | stderr |     "value👉" I will use grounding_dino to help to answer the question. Please wait for a moment.
2023-11-14 14:25:05 | ERROR | stderr |     ^^^^^^^^
2023-11-14 14:25:05 | ERROR | stderr | SyntaxError: invalid syntax

[davidlight2018]

Met the same problem too.

[lxj22]

Same problem here. It seems that the model can not output with the designed format?

[pedramaghazadeh]

The model has issues with outputting the correct format sometimes (Thoughts, actions, and values). In my experience, the issue arises whenever model continues to output random words after successfully answering a question, which ruins the format that is used to extract the tool that needs to be used. Solution In my case, adding "###\n", "\n###", and couple of other strings to the stopping criteria helps with token generation and the output will be usable for the pattern matching regex to extract the correct tool.