Re:ChatGPT Pro
by monkmartinez on Thursday December 05, 2024 @06:44PM
(#42334179)
Memory. I don't know the equation, but its very easy to see when you load a 128k context model at 8K vs 80K. The quant I am running would double VRAM requirements when loading 80K.
[ Parent
| on HN ]
Re:ChatGPT Pro
by thomasfromcdnjs on Thursday December 05, 2024 @09:09PM
(#42335447)
This was my understanding too. Would love more people to chime in on the limits and costs of larger contexts.