Any plan to release a smaller or tiny model?

[lizhiqi49]

Thanks for releasing such an awesome model!

I wonder if you have the plan to further release models with smaller size? The currently released model is of 13B parameters, which can be very expensive for inference and fine-tuning. I think it would emerge many interesting downstream works if a smaller checkpoint was available.

[nitinmukesh]

+1

[4lt3r3go]

instead of a smaller model, perhaps an optimized model like LTX
or 4-bit quantized, as the guys do in this link https://github.com/mit-han-lab/nunchaku