BigGAN_PyTorch/BigGAN.py [44:91]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    arch[256] = {
        "in_channels": [ch * item for item in [16, 16, 8, 8, 4, 2]],
        "out_channels": [ch * item for item in [16, 8, 8, 4, 2, 1]],
        "upsample": [True] * 6,
        "resolution": [8, 16, 32, 64, 128, 256],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 9)
        },
    }
    arch[128] = {
        "in_channels": [ch * item for item in [16, 16, 8, 4, 2]],
        "out_channels": [ch * item for item in [16, 8, 4, 2, 1]],
        "upsample": [True] * 5,
        "resolution": [8, 16, 32, 64, 128],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 8)
        },
    }
    arch[64] = {
        "in_channels": [ch * item for item in [16, 16, 8, 4]],
        "out_channels": [ch * item for item in [16, 8, 4, 2]],
        "upsample": [True] * 4,
        "resolution": [8, 16, 32, 64],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 7)
        },
    }
    arch[32] = {
        "in_channels": [ch * item for item in [4, 4, 4]],
        "out_channels": [ch * item for item in [4, 4, 4]],
        "upsample": [True] * 3,
        "resolution": [8, 16, 32],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 6)
        },
    }

    return arch


class Generator(nn.Module):
    def __init__(
        self,
        G_ch=64,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



BigGAN_PyTorch/BigGANdeep.py [89:136]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    arch[256] = {
        "in_channels": [ch * item for item in [16, 16, 8, 8, 4, 2]],
        "out_channels": [ch * item for item in [16, 8, 8, 4, 2, 1]],
        "upsample": [True] * 6,
        "resolution": [8, 16, 32, 64, 128, 256],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 9)
        },
    }
    arch[128] = {
        "in_channels": [ch * item for item in [16, 16, 8, 4, 2]],
        "out_channels": [ch * item for item in [16, 8, 4, 2, 1]],
        "upsample": [True] * 5,
        "resolution": [8, 16, 32, 64, 128],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 8)
        },
    }
    arch[64] = {
        "in_channels": [ch * item for item in [16, 16, 8, 4]],
        "out_channels": [ch * item for item in [16, 8, 4, 2]],
        "upsample": [True] * 4,
        "resolution": [8, 16, 32, 64],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 7)
        },
    }
    arch[32] = {
        "in_channels": [ch * item for item in [4, 4, 4]],
        "out_channels": [ch * item for item in [4, 4, 4]],
        "upsample": [True] * 3,
        "resolution": [8, 16, 32],
        "attention": {
            2 ** i: (2 ** i in [int(item) for item in attention.split("_")])
            for i in range(3, 6)
        },
    }

    return arch


class Generator(nn.Module):
    def __init__(
        self,
        G_ch=64,
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -