LLaMA-Omni icon indicating copy to clipboard operation
LLaMA-Omni copied to clipboard

Published 20 hours ago • verified publisher icon ictnlp

about train code

Open ali762111 opened this issue 1 year ago • 17 comments

God Work! Is there an open-source plan for training code?

ali762111 avatar Sep 11 '24 04:09 ali762111

Thank you for your interest! Unfortunately, due to some restrictions, we are unable to release the training code at this time. We will update the repository if this changes in the future.

Poeroz avatar Sep 11 '24 13:09 Poeroz

@Poeroz can you atleast give us some guidelines to train and finetune the model.

pashanitw avatar Sep 11 '24 17:09 pashanitw

Good work! Is there an open-source plan for your constructed dataset? Any schedule? Thanks.

moonnee avatar Sep 12 '24 06:09 moonnee

Yes, we plan to release the training data in the future. Please stay tuned.

Poeroz avatar Sep 12 '24 08:09 Poeroz

Is there any esitmated time for data releasing?

MonolithFoundation avatar Sep 12 '24 09:09 MonolithFoundation

+1

kadirnar avatar Sep 17 '24 06:09 kadirnar

Updates? Need this!

yukiarimo avatar Sep 19 '24 21:09 yukiarimo

I'm also interested in your training setup. Especially because since you released Llama-Omni, meta released Llama 3.2 which includes Vision support. It would be great if we could adapt your code to make it really multimodal (TTS+STT + VISION !)

thiswillbeyourgithub avatar Oct 05 '24 11:10 thiswillbeyourgithub

Yes, we plan to release the training data in the future. Please stay tuned.

@Poeroz Hi, is there any progress about releasing the Instruct-S2S dataset?

ZhengRachel avatar Oct 17 '24 12:10 ZhengRachel

Same here

MonolithFoundation avatar Oct 18 '24 06:10 MonolithFoundation

Hi, is any news about release of training code?

risqaliyevds avatar Mar 12 '25 06:03 risqaliyevds

hi @Poeroz what's the word on that training code? Still coming?

sevmardi avatar Mar 26 '25 19:03 sevmardi

You don't need to wait for the training code of the Llama-Omni model. You can use different Omni models. Llama-Omni is not a very good model.

https://github.com/VITA-MLLM/VITA?tab=readme-ov-file#-training

kadirnar avatar Mar 27 '25 11:03 kadirnar

You don't need to wait for the training code of the Llama-Omni model. You can use different Omni models. Llama-Omni is not a very good model.

https://github.com/VITA-MLLM/VITA?tab=readme-ov-file#-training

Thanks for sharing VITA,

If you are aware of other models, please do share.

sevmardi avatar Mar 27 '25 14:03 sevmardi

You don't need to wait for the training code of the Llama-Omni model. You can use different Omni models. Llama-Omni is not a very good model. https://github.com/VITA-MLLM/VITA?tab=readme-ov-file#-training

Thanks for sharing VITA,

If you are aware of other models, please do share.

You can access detailed information about many omni models here.

  • https://github.com/MatthewCYM/VoiceBench?tab=readme-ov-file#awesome-voice-assistants
  • https://github.com/AudioLLMs/Awesome-Audio-LLM/tree/main?tab=readme-ov-file#abbreviations-with-links

kadirnar avatar Mar 27 '25 14:03 kadirnar

Sorry for the long wait. We have released the dataset at this link.

Poeroz avatar May 19 '25 02:05 Poeroz

Sorry for the long wait. We have released the dataset at this link.

Thanks

What about the training code?

sevmardi avatar May 19 '25 05:05 sevmardi