trlx icon indicating copy to clipboard operation
trlx copied to clipboard

Add google's T5 flan to models

Open aashay96 opened this issue 2 years ago • 4 comments

🚀 The feature, motivation, and pitch

GPT-j-neo is currently very limited to the number of languages it supports. Flan has support for more languages. It will allow to train something similar to chatGPT for other languages as well.

Alternatives

https://huggingface.co/google/flan-t5-xxl

Additional context

No response

aashay96 avatar Dec 11 '22 18:12 aashay96

@reshinthadithyan has this working already. He has not merged it yet.

LouisCastricato avatar Dec 12 '22 14:12 LouisCastricato

@reshinthadithyan has this working already. He has not merged it yet.

Which branch? I don't see any PR by @reshinthadithyan

aashay96 avatar Dec 12 '22 19:12 aashay96

any update on this?

mukhal avatar Dec 21 '22 19:12 mukhal

https://github.com/CarperAI/trlx/pull/145

LouisCastricato avatar Dec 21 '22 20:12 LouisCastricato