Post

got bonsai and 27b running decently on my Rx 6700 (10GB VRAM), packages as a nix flake so anyone can run it with minimal work 9b -- 813 tokens/s prefill, 84 tokens//s 27B -- 267 tokens/s prefill, 26 tokens/s github.com/alyraffauf/b...

github.comGitHub - alyraffauf/bonsai-buddy: Run Prism Bonsai 2 9B and 27B locally with Nix, tuned AMD GPU profiles, and OpenCode.Run Prism Bonsai 2 9B and 27B locally with Nix, tuned AMD GPU profiles, and OpenCode. - alyraffauf/bonsai-buddy

7 reposts81 likes

*bonsai 9b and 27b, oops also got decent context windows given size

16K for 27b is still pretty tight but 128k with 9b is actually usable!

I wonder if you can eke out a little more context from bonsai by doing a 1:9 pipeline parallelism between your cpu and gpu, and I also wonder when the accuracy falloff begins from the quantization

(this was a bench I was contemplating writing after noticing that quantized model inaccuracy is less a constant "these tokens are less worth it" and more a "error accumulation builds up, and your first few thousand/ten thousand tokens are actually very usable)

goat. have you messed with Strata or any of its forks?

How much of the fever did you get? Are you doing vulkan vs rocm benchmarks and testing disabling reasoning for more loops? There's an ornith 9b to play with...😈

you could def use this and run 27b with a longer context window.

fun fact I pretend to have your GPU so rocm runs bc its not supported lol