Need guidance on Adding ASR Support to Auto Train UI

🤗AutoTrain
1

Hi Hugging Face Support Team,

I hope you’re doing well!

My name is Aryan Mathur, and I’m currently serving as an NLP Engineering Intern at CognifAI. As part of my internship project, I’ve been tasked with exploring ways to integrate Automatic Speech Recognition (ASR) into the Hugging Face Auto Train platform.

Project Goal:

1:- Add ASR as a task in the Auto Train UI and include ASR-specific models (e.g., Wav2Vec2, Whisper) in the model selection.

Current Issue: ASR is not listed in the Auto Train task section, and ASR models are not available in the model list.

My questions:

  1. Is it technically feasible to add ASR support to the Auto Train UI? If so, could you provide guidance on the steps required?

  2. Are there plans to officially support ASR in Auto Train in the near future?

  3. If ASR is not yet supported, are there documentation or codebase resources I can reference to attempt this integration?

About My Role:
I’m eager to contribute to this project as part of my internship deliverables. Guidance from your team would help me align my efforts with Hugging Face’s architecture and best practices.

Thank you for your time and support!

1 Like
2

Anyway, I think it would be easier to go to the AutoTrainAdvanced github.
I guess the process is to fork, modify the code, and then do PR if necessary.

2

Probably not.

3

It seems that it is possible to train with the Transformers Trainer, so basically, if you can successfully bridge the gap between these two, it should be fine… but it’s easier said than done.