3 ms·
This is the spec for a training node. The inference requires 80GB of VRAM, so significantly less compute.
by vessenes 1y ago
This is the spec for a training node. The inference requires 80GB of VRAM, so significantly less compute.
- andai 1y agoThe default model is ~0.5B params right?