Hi everyone,

Has anyone here fine-tuned facebookresearch/omnilingual-asr for a new language or custom dataset?

I’m trying to set up a full fine-tuning pipeline (data prep → training → evaluation), but the official repo doesn’t provide much detail on adapting the model. If you’ve done it before, could you share:

• Your training workflow
• Any scripts/codebase you used
• Tips on dataset formatting
• Hardware requirements
• Any issues you ran into during fine-tuning

Even a GitHub link or minimal training script would help a lot.
Thanks in advance!