[Bug]: ValueError: Model architectures ['Qwen2AudioForConditionalGeneration'] failed to be inspected. Please check the logs for more details.

[umie0128]

Your current environment

@DarkLight1337
我测试了vllm的 v0.6.4 v0.6.5 v0.6.6 都无法启动qwen2-Aduio
使用的环境都是官方的 docker pull vllm/vllm-openai 对应版本
查看了相关issues 降低Numpy到v1.x 但是官方的Numpy就是1.26.4 并没有效果

Model Input Dumps

Qwen2-Audio-7B-Instruct

🐛 Describe the bug

v0.6.4的报错是：

WARNING 01-14 23:14:06 config.py:1865] Casting torch.bfloat16 to torch.float16.
ERROR 01-14 23:14:12 registry.py:297] Error in inspecting model architecture 'Qwen2AudioForConditionalGeneration'
ERROR 01-14 23:14:12 registry.py:297] Traceback (most recent call last):
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 468, in _run_in_subprocess
ERROR 01-14 23:14:12 registry.py:297] returned.check_returncode()
ERROR 01-14 23:14:12 registry.py:297] File "/usr/lib/python3.12/subprocess.py", line 502, in check_returncode
ERROR 01-14 23:14:12 registry.py:297] raise CalledProcessError(self.returncode, self.args, self.stdout,
ERROR 01-14 23:14:12 registry.py:297] subprocess.CalledProcessError: Command '['/usr/bin/python3', '-m', 'vllm.model_executor.models.registry']' returned non-zero exit status 1.
ERROR 01-14 23:14:12 registry.py:297]
ERROR 01-14 23:14:12 registry.py:297] The above exception was the direct cause of the following exception:
ERROR 01-14 23:14:12 registry.py:297]
ERROR 01-14 23:14:12 registry.py:297] Traceback (most recent call last):
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 295, in _try_inspect_model_cls
ERROR 01-14 23:14:12 registry.py:297] return model.inspect_model_cls()
ERROR 01-14 23:14:12 registry.py:297] ^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 257, in inspect_model_cls
ERROR 01-14 23:14:12 registry.py:297] return _run_in_subprocess(
ERROR 01-14 23:14:12 registry.py:297] ^^^^^^^^^^^^^^^^^^^
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 471, in _run_in_subprocess
ERROR 01-14 23:14:12 registry.py:297] raise RuntimeError(f"Error raised in subprocess:\n"
ERROR 01-14 23:14:12 registry.py:297] RuntimeError: Error raised in subprocess:
ERROR 01-14 23:14:12 registry.py:297] :128: RuntimeWarning: 'vllm.model_executor.models.registry' found in sys.modules after import of package 'vllm.model_executor.models', but prior to execution of 'vllm.model_executor.models.registry'; this may result in unpredictable behaviour
ERROR 01-14 23:14:12 registry.py:297] Traceback (most recent call last):
ERROR 01-14 23:14:12 registry.py:297] File "", line 198, in _run_module_as_main
ERROR 01-14 23:14:12 registry.py:297] File "", line 88, in _run_code
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 492, in
ERROR 01-14 23:14:12 registry.py:297] _run()
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 485, in _run
ERROR 01-14 23:14:12 registry.py:297] result = fn()
ERROR 01-14 23:14:12 registry.py:297] ^^^^
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 258, in
ERROR 01-14 23:14:12 registry.py:297] lambda: _ModelInfo.from_model_cls(self.load_model_cls()))
ERROR 01-14 23:14:12 registry.py:297] ^^^^^^^^^^^^^^^^^^^^^
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 261, in load_model_cls
ERROR 01-14 23:14:12 registry.py:297] mod = importlib.import_module(self.module_name)
ERROR 01-14 23:14:12 registry.py:297] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 01-14 23:14:12 registry.py:297] File "/usr/lib/python3.12/importlib/init.py", line 90, in import_module
ERROR 01-14 23:14:12 registry.py:297] return _bootstrap._gcd_import(name[level:], package, level)
ERROR 01-14 23:14:12 registry.py:297] ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
ERROR 01-14 23:14:12 registry.py:297] File "", line 1387, in _gcd_import
ERROR 01-14 23:14:12 registry.py:297] File "", line 1360, in _find_and_load
ERROR 01-14 23:14:12 registry.py:297] File "", line 1331, in _find_and_load_unlocked
ERROR 01-14 23:14:12 registry.py:297] File "", line 935, in _load_unlocked
ERROR 01-14 23:14:12 registry.py:297] File "", line 995, in exec_module
ERROR 01-14 23:14:12 registry.py:297] File "", line 488, in _call_with_frames_removed
ERROR 01-14 23:14:12 registry.py:297] File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2_audio.py", line 25, in
ERROR 01-14 23:14:12 registry.py:297] import librosa
ERROR 01-14 23:14:12 registry.py:297] ModuleNotFoundError: No module named 'librosa'
ERROR 01-14 23:14:12 registry.py:297]
Traceback (most recent call last):
File "/workspace/online_server.py", line 357, in
engine = AsyncLLMEngine.from_engine_args(engine_args, usage_context=UsageContext.OPENAI_API_SERVER) # 不传默认是UsageContext.ENGINE_CONTEXT
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/engine/async_llm_engine.py", line 683, in from_engine_args
engine_config = engine_args.create_engine_config()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/engine/arg_utils.py", line 959, in create_engine_config
model_config = self.create_model_config()
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/engine/arg_utils.py", line 891, in create_model_config
return ModelConfig(
^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/config.py", line 251, in init
self.multimodal_config = self._init_multimodal_config(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/config.py", line 277, in _init_multimodal_config
if ModelRegistry.is_multimodal_model(architectures):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 422, in is_multimodal_model
return self.inspect_model_cls(architectures).supports_multimodal
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 391, in inspect_model_cls
return self._raise_for_unsupported(architectures)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/registry.py", line 348, in _raise_for_unsupported
raise ValueError(
ValueError: Model architectures ['Qwen2AudioForConditionalGeneration'] failed to be inspected. Please check the logs for more details.

v0.6.6的报错是：

INFO 01-14 23:11:06 model_runner.py:1099] Loading model weights took 15.6454 GB
[rank0]: Traceback (most recent call last):
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/utils.py", line 1588, in getattr
[rank0]: importlib.import_module(self.name)
[rank0]: File "/usr/lib/python3.12/importlib/init.py", line 90, in import_module
[rank0]: return _bootstrap._gcd_import(name[level:], package, level)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "", line 1387, in _gcd_import
[rank0]: File "", line 1360, in _find_and_load
[rank0]: File "", line 1324, in _find_and_load_unlocked
[rank0]: ModuleNotFoundError: No module named 'librosa'

[rank0]: The above exception was the direct cause of the following exception:

[rank0]: Traceback (most recent call last):
[rank0]: File "/workspace/online_server.py", line 357, in
[rank0]: engine = AsyncLLMEngine.from_engine_args(engine_args, usage_context=UsageContext.OPENAI_API_SERVER) # 不传默认是UsageContext.ENGINE_CONTEXT
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/engine/async_llm_engine.py", line 707, in from_engine_args
[rank0]: engine = cls(
[rank0]: ^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/engine/async_llm_engine.py", line 594, in init
[rank0]: self.engine = self._engine_class(*args, **kwargs)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/engine/async_llm_engine.py", line 267, in init
[rank0]: super().init(*args, **kwargs)
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/engine/llm_engine.py", line 276, in init
[rank0]: self._initialize_kv_caches()
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/engine/llm_engine.py", line 416, in _initialize_kv_caches
[rank0]: self.model_executor.determine_num_available_blocks())
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/executor/gpu_executor.py", line 68, in determine_num_available_blocks
[rank0]: return self.driver_worker.determine_num_available_blocks()
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank0]: return func(*args, **kwargs)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/worker/worker.py", line 202, in determine_num_available_blocks
[rank0]: self.model_runner.profile_run()
[rank0]: File "/usr/local/lib/python3.12/dist-packages/torch/utils/_contextlib.py", line 116, in decorate_context
[rank0]: return func(*args, **kwargs)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/worker/model_runner.py", line 1291, in profile_run
[rank0]: .dummy_data_for_profiling(self.model_config,
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/inputs/registry.py", line 339, in dummy_data_for_profiling
[rank0]: dummy_data = processor.get_dummy_data(seq_len, mm_counts,
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/multimodal/processing.py", line 858, in get_dummy_data
[rank0]: mm_inputs = self.apply(*processor_inputs)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/multimodal/processing.py", line 795, in apply
[rank0]: hf_inputs = self._apply_hf_processor(prompt_text, mm_items,
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/multimodal/processing.py", line 704, in _apply_hf_processor
[rank0]: processor_data, passthrough_data = self._get_processor_data(mm_items)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/model_executor/models/qwen2_audio.py", line 103, in _get_processor_data
[rank0]: mm_items.resample_audios(feature_extractor.sampling_rate)
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/multimodal/processing.py", line 302, in resample_audios
[rank0]: audio = resample_audio(audio, orig_sr=sr, target_sr=new_sr)
[rank0]: ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/multimodal/audio.py", line 41, in resample_audio
[rank0]: return librosa.resample(audio, orig_sr=orig_sr, target_sr=target_sr)
[rank0]: ^^^^^^^^^^^^^^^^
[rank0]: File "/usr/local/lib/python3.12/dist-packages/vllm/utils.py", line 1593, in getattr
[rank0]: raise ImportError(msg) from exc
[rank0]: ImportError: Please install vllm[audio] for audio support

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.