Add FunASR/SenseVoice models as supported ASR in SlamKit

[LauraGPT]

Hi! Great toolkit for efficient SpeechLM training.

Would you consider supporting FunASR models (Paraformer, SenseVoice) in SlamKit?

Relevant models

• Fun-ASR-Nano — Audio encoder + Qwen2.5-0.5B LLM, end-to-end speech LM
• SenseVoice — Multi-task speech model (ASR + emotion + events), non-autoregressive
• Paraformer — Non-autoregressive ASR with CTC/attention hybrid

Why integrate?

• Pre-trained audio encoders from FunASR could serve as speech encoders for SpeechLM training
• SenseVoice's multi-task architecture is relevant for multi-task SpeechLM research
• Fun-ASR-Nano demonstrates encoder-LLM fusion for speech understanding

References

• Fun-ASR-Nano: https://github.com/FunAudioLLM/Fun-ASR (1.2K stars)
• SenseVoice: https://github.com/FunAudioLLM/SenseVoice (8K+ stars)
• FunASR: https://github.com/modelscope/FunASR (16K+ stars)
• Paper: https://arxiv.org/abs/2407.04051