Network with 1 layer possible?

[acse-pms122]

Hi, is it possible to create a tinycuda neural net with simply one dense layer. I get an error when I set n_hidden_layers to 0. But is there some kind of workaround for this?

Thanks in advance!

[amitabe]

With CutlassMLP this works for me:

import tinycudann as tcnn

config = {
    "otype": "CutlassMLP",
    "activation": "None",
    "output_activation": "LeakyReLU",
    "n_neurons": 0,
    "n_hidden_layers": 0,
}
proj = tcnn.Network(2, 256, proj_config)