6 ms·
Inference Arena – new benchmark of local inference and training
- Alexzoofficial 6mo ago[flagged]
- kvark 6mo agoThank you for support! Part of the problem with timing variety is frameworks not always picking the right gpu/backend. If you want to inspect or tweak the setup, be my guest at https://github.com/kvark/inferena https://github.com/kvark/inferena