-
exllama
A more memory-efficient rewrite of the HF transformers implementation of Llama for use with quantized weights.
-
Scout Monitoring
Free Django app performance insights with Scout Monitoring. Get Scout setup in minutes, and let us sweat the small stuff. A couple lines in settings.py is all you need to start monitoring your apps. Sign up for our free tier today.
> We serve Llama on 2 80-GB A100 GPUs, as that is the minumum required to fit Llama in memory (with 16-bit precision)
Well there is your problem.
LLaMA quantized to 4 bits fits in 40GB. And it gets similar throughput split between dual consumer GPUs, which likely means better throughput on a single 40GB A100 (or a cheaper 48GB Pro GPU)
https://github.com/turboderp/exllama#dual-gpu-results
Also, I'm not sure which model was tested, but Llama 70B chat should have better performance than the base model if the prompting syntax is right. That was only reverse engineered from the Meta demo implementation recently.