chenwydj / darts_evaluation Goto Github PK

I got this darts genotype from TENAS(as below)，I wonder why "Genotype" only contains information about node 0 and node 1. Is this result format correct? In my opinion, I think these two tensor matrix are correct, do I need to translate these two tensor matrix into type "Genotype" to evaluate this architecture? Looking forward to your reply, thank you very much.

End: Genotype(normal=[('sep_conv_5x5', 0), ('sep_conv_5x5', 1), ('sep_conv_5x5', 0), ('sep_conv_5x5', 1), ('sep_conv_5x5', 0), ('sep_conv_5x5', 1), ('sep_conv_5x5', 0), ('sep_conv_5x5', 1)], normal_concat=[2, 3, 4, 5], reduce=[('dil_conv_5x5', 0), ('sep_conv_3x3', 1), ('dil_conv_5x5', 0), ('sep_conv_3x3', 1), ('dil_conv_5x5', 0), ('sep_conv_3x3', 1), ('dil_conv_5x5', 0), ('sep_conv_3x3', 1)], reduce_concat=[2, 3, 4, 5])
operators remaining (1s) and prunned (0s)
tensor([[0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 1],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0]], device='cuda:0', dtype=torch.int32)
tensor([[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 1, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 1, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 1, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 1, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0]], device='cuda:0', dtype=torch.int32)