BUG: ChatGLM Parameter Switch

[Bojun-Feng]

Describe the bug

ChatGLM fails to start, seems like two parameters model_path and quantization were passed in reverse.

Additional context

Full IPython History

In [1]: from xinference.client import Client
   ...: 
   ...: client = Client("http://localhost:9997")
   ...: model_uid = client.launch_model(model_name="chatglm2")
   ...: model = client.get_model(model_uid)
---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
Cell In[1], line 4
      1 from xinference.client import Client
      3 client = Client("http://localhost:9997")
----> 4 model_uid = client.launch_model(model_name="chatglm2")
      5 model = client.get_model(model_uid)

File ~/Desktop/cs/inference/xinference/client.py:282, in Client.launch_model(self, model_name, model_size_in_billions, model_format, quantization, **kwargs)
    272 model_uid = self.gen_model_uid()
    274 coro = self._supervisor_ref.launch_builtin_model(
    275     model_uid=model_uid,
    276     model_name=model_name,
   (...)
    280     **kwargs,
    281 )
--> 282 self._isolation.call(coro)
    284 return model_uid

File ~/Desktop/cs/inference/xinference/isolation.py:44, in Isolation.call(self, coro)
     42 def call(self, coro: Coroutine) -> Any:
     43     fut = asyncio.run_coroutine_threadsafe(coro, self._loop)
---> 44     return fut.result()

File ~/anaconda3/lib/python3.11/concurrent/futures/_base.py:456, in Future.result(self, timeout)
    454     raise CancelledError()
    455 elif self._state == FINISHED:
--> 456     return self.__get_result()
    457 else:
    458     raise TimeoutError()

File ~/anaconda3/lib/python3.11/concurrent/futures/_base.py:401, in Future.__get_result(self)
    399 if self._exception:
    400     try:
--> 401         raise self._exception
    402     finally:
    403         # Break a reference cycle with the exception in self._exception
    404         self = None

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/context.py:227, in IndigenActorContext.send(self, actor_ref, message, wait_response, profiling_context)
    225 if wait_response:
    226     result = await self._wait(future, actor_ref.address, send_message)  # type: ignore
--> 227     return self._process_result_message(result)
    228 else:
    229     return future

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/context.py:102, in IndigenActorContext._process_result_message(message)
    100     return message.result
    101 else:
--> 102     raise message.as_instanceof_cause()

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/pool.py:657, in send()
    655         raise ActorNotExist(f"Actor {actor_id} does not exist")
    656     coro = self._actors[actor_id].__on_receive__(message.content)
--> 657     result = await self._run_coro(message.message_id, coro)
    658     processor.result = ResultMessage(
    659         message.message_id,
    660         result,
    661         protocol=message.protocol,
    662         profiling_context=message.profiling_context,
    663     )
    664 return processor.result

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/pool.py:368, in _run_coro()
    366 self._process_messages[message_id] = asyncio.tasks.current_task()
    367 try:
--> 368     return await coro
    369 finally:
    370     self._process_messages.pop(message_id, None)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/api.py:306, in __on_receive__()
    297 async def __on_receive__(self, message: Tuple[Any]):
    298     """
    299     Handle message from other actors and dispatch them to user methods
    300 
   (...)
    304         Message shall be (method_name,) + args + (kwargs,)
    305     """
--> 306     return await super().__on_receive__(message)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:527, in __on_receive__()
    525     debug_logger.exception('Got unhandled error when handling message %.500r '
    526                            'in actor %s at %s', message, self.uid, self.address)
--> 527 raise ex
    528 
    529 

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:519, in xoscar.core._BaseActor.__on_receive__()
    517         raise ValueError(f'call_method {call_method} not valid')
    518 
--> 519     return await self._handle_actor_result(result)
    520 except Exception as ex:
    521     if _log_unhandled_errors:

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:404, in _handle_actor_result()
    402 # asyncio.wait as it introduces much overhead
    403 if len(coros) == 1:
--> 404     task_result = await coros[0]
    405     if extract_tuple:
    406         result = task_result

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:447, in _run_actor_async_generator()
    445 res = None
    446 while True:
--> 447     async with self._lock:
    448         with debug_async_timeout('actor_lock_timeout',
    449                                  'async_generator %r hold lock timeout', gen):

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:448, in xoscar.core._BaseActor._run_actor_async_generator()
    446 while True:
    447     async with self._lock:
--> 448         with debug_async_timeout('actor_lock_timeout',
    449                                  'async_generator %r hold lock timeout', gen):
    450             if not is_exception:

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:453, in xoscar.core._BaseActor._run_actor_async_generator()
    451             res = await gen.asend(res)
    452         else:
--> 453             res = await gen.athrow(*res)
    454 try:
    455     if _log_cycle_send:

File ~/Desktop/cs/inference/xinference/core/service.py:119, in launch_builtin_model()
    116     raise ValueError(f"Model is already in the model list, uid: {model_uid}")
    118 worker_ref = await self._choose_worker()
--> 119 model_ref = yield worker_ref.launch_builtin_model(
    120     model_uid=model_uid,
    121     model_name=model_name,
    122     model_size_in_billions=model_size_in_billions,
    123     model_format=model_format,
    124     quantization=quantization,
    125     **kwargs,
    126 )
    127 # TODO: not protected.
    128 self._model_uid_to_worker[model_uid] = worker_ref

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:458, in xoscar.core._BaseActor._run_actor_async_generator()
    456         message_trace = pop_message_trace()
    457 
--> 458     res = await self._handle_actor_result(res)
    459     is_exception = False
    460 except:

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:378, in _handle_actor_result()
    376 
    377         if inspect.isawaitable(result):
--> 378             result = await result
    379         elif is_async_generator(result):
    380             result = (result,)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:266, in __pyx_actor_method_wrapper()
    264 
    265 async def __pyx_actor_method_wrapper(method, result_handler, lock, args, kwargs):
--> 266     async with lock:
    267         result = method(*args, **kwargs)
    268         if asyncio.iscoroutine(result):

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:269, in xoscar.core.__pyx_actor_method_wrapper()
    267     result = method(*args, **kwargs)
    268     if asyncio.iscoroutine(result):
--> 269         result = await result
    270 return await result_handler(result)
    271 

File ~/Desktop/cs/inference/xinference/core/service.py:44, in wrapped()
     42 logger.debug(f"Enter {func.__name__}, args: {args}, kwargs: {kwargs}")
     43 start = time.time()
---> 44 ret = await func(*args, **kwargs)
     45 logger.debug(
     46     f"Leave {func.__name__}, elapsed time: {int(time.time() - start)} ms"
     47 )
     48 return ret

File ~/Desktop/cs/inference/xinference/core/service.py:324, in launch_builtin_model()
    320 subpool_address = self._choose_subpool()
    321 model_ref = await xo.create_actor(
    322     ModelActor, address=subpool_address, uid=model_uid, model=model
    323 )
--> 324 await model_ref.load()
    325 self._model_uid_to_model[model_uid] = model_ref
    326 self._model_uid_to_model_spec[model_uid] = (llm_family, llm_spec, quantization)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/context.py:227, in send()
    225 if wait_response:
    226     result = await self._wait(future, actor_ref.address, send_message)  # type: ignore
--> 227     return self._process_result_message(result)
    228 else:
    229     return future

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/context.py:102, in _process_result_message()
    100     return message.result
    101 else:
--> 102     raise message.as_instanceof_cause()

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/pool.py:657, in send()
    655         raise ActorNotExist(f"Actor {actor_id} does not exist")
    656     coro = self._actors[actor_id].__on_receive__(message.content)
--> 657     result = await self._run_coro(message.message_id, coro)
    658     processor.result = ResultMessage(
    659         message.message_id,
    660         result,
    661         protocol=message.protocol,
    662         profiling_context=message.profiling_context,
    663     )
    664 return processor.result

File ~/anaconda3/lib/python3.11/site-packages/xoscar/backends/pool.py:368, in _run_coro()
    366 self._process_messages[message_id] = asyncio.tasks.current_task()
    367 try:
--> 368     return await coro
    369 finally:
    370     self._process_messages.pop(message_id, None)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/api.py:306, in __on_receive__()
    297 async def __on_receive__(self, message: Tuple[Any]):
    298     """
    299     Handle message from other actors and dispatch them to user methods
    300 
   (...)
    304         Message shall be (method_name,) + args + (kwargs,)
    305     """
--> 306     return await super().__on_receive__(message)

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:527, in __on_receive__()
    525     debug_logger.exception('Got unhandled error when handling message %.500r '
    526                            'in actor %s at %s', message, self.uid, self.address)
--> 527 raise ex
    528 
    529 

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:497, in xoscar.core._BaseActor.__on_receive__()
    495 if call_method == CALL_METHOD_DEFAULT:
    496     func = getattr(self, method)
--> 497     async with self._lock:
    498         with debug_async_timeout('actor_lock_timeout',
    499                                  "Method %s of actor %s hold lock timeout.",

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:498, in xoscar.core._BaseActor.__on_receive__()
    496 func = getattr(self, method)
    497 async with self._lock:
--> 498     with debug_async_timeout('actor_lock_timeout',
    499                              "Method %s of actor %s hold lock timeout.",
    500                              method, self.uid):

File ~/anaconda3/lib/python3.11/site-packages/xoscar/core.pyx:501, in xoscar.core._BaseActor.__on_receive__()
    499                      "Method %s of actor %s hold lock timeout.",
    500                      method, self.uid):
--> 501 result = func(*args, **kwargs)
    502 if asyncio.iscoroutine(result):
    503     result = await result

File ~/Desktop/cs/inference/xinference/core/model.py:86, in load()
     85 def load(self):
---> 86     self._model.load()

File ~/Desktop/cs/inference/xinference/model/llm/ggml/chatglm.py:84, in load()
     74     installation_guide = [
     75         "Please make sure 'chatglm_cpp' is installed. ",
     76         "You can install it by running the following command in the terminal:\n",
   (...)
     79         "https://github.com/li-plus/chatglm.cpp",
     80     ]
     82     raise ImportError(f"{error_message}\n\n{''.join(installation_guide)}")
---> 84 self._llm = chatglm_cpp.Pipeline(Path(self.model_path))

File ~/anaconda3/lib/python3.11/site-packages/chatglm_cpp/__init__.py:13, in __init__()
     11 def __init__(self, model_path: os.PathLike) -> None:
     12     model_path = Path(model_path)
---> 13     super().__init__(str(model_path))

RuntimeError: [address=localhost:62884, pid=51919] /Users/fengbojun/Desktop/cs/chatglm.cpp/chatglm.cpp:250 check failed (fd > 0) cannot open file q4_0: No such file or directory