undefined | Better HN

0 pointsvessenes8mo ago0 comments

This is the spec for a training node. The inference requires 80GB of VRAM, so significantly less compute.

0 comments

1 comments · 1 top-level

andai8mo ago

The default model is ~0.5B params right?

j / k navigate · click thread line to collapse