FunASR-Nano

https://modelscope.cn/models/FunAudioLLM/Fun-ASR-Nano-2512/file/view/master/config.yam

Audio encoder is SenseVoiceEncoderSmall.

LLM is Qwen3-0.6B, from transformers.AutoModelForCausalLM. See https://huggingface.co/Qwen/Qwen3-0.6B

Audio adaptor is Transformer from funasr/models/transformer

[
  [
    {'role': 'system', 'content': 'You are a helpful assistant.'},
    {'role': 'user', 'content': '语音转写：<|startofspeech|>!/root/.cache/huggingface/hub/models--FunAudioLLM--Fun-ASR-Nano-2512/snapshots/6d5631d3240449810e6dbcf92c9af0ca063e12cb/example/zh.mp3<|endofspeech|>'},
    {'role': 'assistant', 'content': 'null'}
  ]
]