Running LLMs locally to create custom dialogue for games is still years away.
Running LLMs locally to create custom dialogue for games is still years away.
I had think about this, you need a small LLM for a game, you do not need it to know about movies, music bands, history, coding and all the text on the internet. I am thinking we need a small model similar to phi-2 , trained only on basic stuff then you would train it on the game world lore. Then the game would also use some "simpler" graphics( we had good looking game graphics decades ago so I do not think you could be limited to text adventures or 2D graphics, just you need some simpler and optimized graphics, it is always interesting when someone shows his unreal demo that uses most RAM/VRAM for a simple demo level then a giant game like GTA5)
Or even better look at something like phi-2. It's likely to go even lower. I'm sure there are people here who can detail more specifics.
Its 1/24 and we're already there more or less.
What do you mean? Most gamers do have an nvidia GPU.
Edit: unless you talk about mobile gamers, and not PC gamers?
https://store.steampowered.com/hwsurvey/Steam-Hardware-Softw...