Llama 3.2 3B Instruct Q40 Benchmarks, 1x, 2x, 4x, 8x Raspberry Pi 4B 8GB #124
githuba9f5404
started this conversation in
Results
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Running on this hardware setup, overclocked with a clock speed of 2000 MHZ (standard is 1500 MHZ). All benchmarks run with
--steps 64
and--prompt "The Eiffel Tower is"
. Output was all reasonable. Model converted to q40, run with q80 buffer.1 x Node
2 x Nodes
4 x Nodes
8 x Nodes
Beta Was this translation helpful? Give feedback.
All reactions