Support falcon models

[bitsnaps]

Feature request

I'm not sure if it's already there, but since GPT4ll is mentioned at Falcon-7B-Instruct I think it's not gonna be hard to be implemented, right?

Motivation

A better performance, accuracy since falcon is outperforming most other models on NLP benchmarks.

Your contribution

I'm not sure if I can, neither if I have permissions!

[mvenditto]

I may be wrong, but it's probably related to if/when the model will be supported in [llama.cpp](ggerganov/llama.cpp#1602

Update:
Was indeed wrong :)

[AndriyMulyar]

We don't need llama.cpp to support falcon for us to support it. Notice we support MPT and GPTJ :)
It's being worked on.