Can someone tell me if I’m being stupid but for the mamba 3 paper do they make available the trained model they provide all their benchmark results for? Cause I can’t see it on HF anywhere and the demo they give has you just passing rand suggesting to me they haven’t provided the trained version which seems odd.