Hacker News new | ask | show | jobs
by mluo 497 days ago
Try bfloat16! We have a bug where the model was saved as fp32.
1 comments
simonw 497 days ago
I just tried it with this 3.6GB F16 model:

  ollama run hf.co/bartowski/agentica-org_DeepScaleR-1.5B-Preview-GGUF:F16
And this time it didn't get confused with the tokenization of strawberry! https://gist.github.com/simonw/9e79f96d69f10bc7ba540c87ea0e8...
link
mluo 497 days ago
Nice, very glad to see it works! Small models are very sensitive to the dtype :(
link