Support for Gemma 7B

[1Krypt0]

Hi!

Are you considering adding support for Gemma 7B? It seems that it would be a great addition to the set of available models.

[joecummings]

We're definitely interested in adding more models!

As I understand it, the 2B and 7B architectures are roughly the same (just different sizes for the parameters). If you'd be interested in adding the model yourself, we'd gladly stamp it :) It should be fairly simple, just adding:

1. A new gemma_7b function in the gemma/_model_builders.py file with the appropriate sizes
2. A screenshot or attached W&B log showing that the model learns from a simple alpaca fine-tuning
3. Updating our models docs to show we now support it!

LMK if you have any questions or if this is something you'd like to take on.

[Optimox]

I am on it!