The Nano only has 4GB VRAM and DS-R1 is 671B FP8 parameters (equivalent to 671GB model size).

You need something with about 800GB to run the full model with context. You'd still need 400GB to even run a half-sized Q4 quant of R1, so there is no reasonable way that it would work.