llama3 slow generation

[adriiglz]

I've been trying to integrate llama3 messages after game ending for giving users feedback in some original way, but text generation is very slow and makes user experience a bit bad.
I presume in our ARM VM it would run even slower.
If anybody has any idea to make llama3 run faster I will continue developing this feature, but I can't integrate it until then.

[andrrsin]

I believe we should focus on other important things, It's a good idea but we are issing key features of the project