为什么 Qwen2.5-Omni-7B 官方教程都报错 Cannot import available module of Qwen2_5OmniModel in modelscope ？

Question

官方教程：https://modelscope.cn/models/Qwen/Qwen2.5-Omni-7B

且安装了最新的 modelscope 版本

╰─➤  pip install modelscope --upgrade                                                                                                                                                                                                                                                   1 ↵
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: modelscope in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (1.17.1)
Collecting modelscope
  Using cached https://pypi.tuna.tsinghua.edu.cn/packages/1b/8b/a3a6a5b3afe89517f513be7257193ba5aed0b164c42e4453ffdd150729d6/modelscope-1.24.1-py3-none-any.whl (5.9 MB)
Requirement already satisfied: requests>=2.25 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from modelscope) (2.32.3)
Requirement already satisfied: tqdm>=4.64.0 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from modelscope) (4.66.5)
Requirement already satisfied: urllib3>=1.26 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from modelscope) (2.2.2)
Requirement already satisfied: charset-normalizer<4,>=2 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from requests>=2.25->modelscope) (3.3.2)
Requirement already satisfied: idna<4,>=2.5 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from requests>=2.25->modelscope) (3.8)
Requirement already satisfied: certifi>=2017.4.17 in /home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages (from requests>=2.25->modelscope) (2024.7.4)
Installing collected packages: modelscope
  Attempting uninstall: modelscope
    Found existing installation: modelscope 1.17.1
    Uninstalling modelscope-1.17.1:
      Successfully uninstalled modelscope-1.17.1
Successfully installed modelscope-1.24.1

[notice] A new release of pip is available: 24.2 -> 25.0.1
[notice] To update, run: pip install --upgrade pip

运行官方文档中的 demo 代码

import soundfile as sf

from modelscope import Qwen2_5OmniModel, Qwen2_5OmniProcessor
from qwen_omni_utils import process_mm_info

# default: Load the model on the available device(s)
model = Qwen2_5OmniModel.from_pretrained("Qwen/Qwen2.5-Omni-7B", torch_dtype="auto", device_map="auto")

# We recommend enabling flash_attention_2 for better acceleration and memory saving.
# model = Qwen2_5OmniModel.from_pretrained(
#     "Qwen/Qwen2.5-Omni-7B",
#     torch_dtype="auto",
#     device_map="auto",
#     attn_implementation="flash_attention_2",
# )

processor = Qwen2_5OmniProcessor.from_pretrained("Qwen/Qwen2.5-Omni-7B")

conversation = [
    {
        "role": "system",
        "content": "You are Qwen, a virtual human developed by the Qwen Team, Alibaba Group, capable of perceiving auditory and visual inputs, as well as generating text and speech.",
    },
    {
        "role": "user",
        "content": [
            {"type": "video", "video": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2.5-Omni/draw.mp4"},
        ],
    },
]

# set use audio in video
USE_AUDIO_IN_VIDEO = True

# Preparation for inference
text = processor.apply_chat_template(conversation, add_generation_prompt=True, tokenize=False)
audios, images, videos = process_mm_info(conversation, use_audio_in_video=USE_AUDIO_IN_VIDEO)
inputs = processor(text=text, audios=audios, images=images, videos=videos, return_tensors="pt", padding=True, use_audio_in_video=USE_AUDIO_IN_VIDEO)
inputs = inputs.to(model.device).to(model.dtype)

# Inference: Generation of the output text and audio
text_ids, audio = model.generate(**inputs, use_audio_in_video=USE_AUDIO_IN_VIDEO)

text = processor.batch_decode(text_ids, skip_special_tokens=True, clean_up_tokenization_spaces=False)
print(text)
sf.write(
    "output.wav",
    audio.reshape(-1).detach().cpu().numpy(),
    samplerate=24000,
)

但是报错了

─➤  python Qwen2_5-Omni-7B.py       
Traceback (most recent call last):
  File "/home/pon/code/me/modelscope_example/Qwen2_5-Omni-7B.py", line 3, in <module>
    from modelscope import Qwen2_5OmniModel, Qwen2_5OmniProcessor
  File "/home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages/modelscope/utils/import_utils.py", line 440, in __getattr__
    value = self._extra_import_func(name)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/pon/.local/share/virtualenvs/modelscope_example-DACykz4b/lib/python3.11/site-packages/modelscope/__init__.py", line 134, in try_import_from_hf
    raise ImportError(
ImportError: Cannot import available module of Qwen2_5OmniModel in modelscope, or related packages(['transformers', 'peft', 'diffusers'])

让我很是无语

Answer 1

已解决。需要源码安装 transformers 以及最新的 modelscope：

pip install git+https://github.com/huggingface/transformers@3a1ead0aabed473eafe527915eea8c197d424356
pip install modelscope -U

参考来源：

• https://github.com/orgs/modelscope/discussions/1295
• https://github.com/modelscope/modelscope/issues/1296