Support for MiMo-VL-7B

[ChangtiWu]

If I want to evaluate MiMo-VL-7B-SFT, do I just need to use the Qwen2.5-VL model configuration?
I tried it once and found that the evaluation was very slow and the final result was 0, it must be something wrong with it.
Will it support MiMo？

[Luodian]

hi Can you check the repo posted by Mimo's team?

https://github.com/XiaomiMiMo/lmms-eval

Or you could use vllm as inference backend, since it supports most existing models (as long Mimo supports it).

[mathCrazyy]

hi Can you check the repo posted by Mimo's team?

https://github.com/XiaomiMiMo/lmms-eval

Or you could use vllm as inference backend, since it supports most existing models (as long Mimo supports it).

#751 哈喽，我尝试使用vllm 作为后端进行评测，但是发现了这个情况，具体是什么原因呢

[TobiasLee]

If I want to evaluate MiMo-VL-7B-SFT, do I just need to use the Qwen2.5-VL model configuration? I tried it once and found that the evaluation was very slow and the final result was 0, it must be something wrong with it. Will it support MiMo？

Hi, thanks for your interest. Please adopt our official evaluation framework for reproducible evaluation of MiMo-VL. https://github.com/XiaomiMiMo/lmms-eval

[Luodian]

Closing due to inactivity (190 days). Feel free to reopen if still relevant.