mt/thesis/third_party/PlotNeuralNet/deepsad/arch_ef_decoder.py

# subter_lenet_arch.py
# Requires running from inside the PlotNeuralNet repo, like: python3 ../subter_lenet_arch.py
import sys, argparse

sys.path.append("../")  # import pycore from repo root

from pycore.tikzeng import *

parser = argparse.ArgumentParser()
parser.add_argument("--rep_dim", type=int, default=1024, help="latent size for FC")
args = parser.parse_args()
REP = int(args.rep_dim)

# Visual scales so the huge width doesn't dominate the figure
H32, H16, H8, H1 = 26, 18, 12, 1
D2048, D1024, D512, D256, D128, D1 = 52, 36, 24, 12, 6, 1
W1, W4, W8, W16, W32 = 1, 2, 2, 4, 8


arch = [
    to_head(".."),
    to_cor(),
    to_begin(),
    to_Conv(
        "latent",
        n_filer="",
        s_filer="latent dim",
        offset="(2,0,0)",
        to="(0,0,0)",
        height=H8 * 1.6,
        depth=D1,
        width=W1,
        caption=f"Latent Space",
    ),
    # to_connection("fc1", "latent"),
    # --------------------------- DECODER ---------------------------
    # FC back to 16384
    to_fc(
        "fc3",
        n_filer="{{8×128×8}}",
        offset="(2,0,0)",
        to="(latent-east)",
        height=H1,
        depth=D512,
        width=W1,
        caption=f"FC",
    ),
    to_Conv(
        "unsqueeze",
        s_filer="",
        n_filer=32,
        offset="(2,0,0)",
        to="(fc3-east)",
        height=H8,
        depth=D128,
        width=W32,
        caption="unsqueeze",
    ),
    # to_connection("latent", "fc3"),
    # Reshape to 4×8×512
    to_UnPool(
        "up1",
        offset="(2,0,0)",
        to="(unsqueeze-east)",
        height=H16,
        depth=D256,
        width=W32,
        caption="",
    ),
    to_Conv(
        "dwdeconv1",
        s_filer="",
        n_filer=1,
        offset="(0,0,0)",
        to="(up1-east)",
        height=H16,
        depth=D256,
        width=W1,
        caption="deconv1",
    ),
    to_Conv(
        "dwdeconv2",
        s_filer="{{256×16}}",
        n_filer=32,
        offset="(0,0,0)",
        to="(dwdeconv1-east)",
        height=H16,
        depth=D256,
        width=W32,
        caption="",
    ),
    to_UnPool(
        "up2",
        offset="(2,0,0)",
        to="(dwdeconv2-east)",
        height=H16,
        depth=D1024,
        width=W32,
        caption="",
    ),
    to_Conv(
        "dwdeconv3",
        s_filer="",
        n_filer=1,
        offset="(0,0,0)",
        to="(up2-east)",
        height=H16,
        depth=D1024,
        width=W1,
        caption="deconv2",
    ),
    to_Conv(
        "dwdeconv4",
        s_filer="{{1024×16}}",
        n_filer=16,
        offset="(0,0,0)",
        to="(dwdeconv3-east)",
        height=H16,
        depth=D1024,
        width=W16,
        caption="",
    ),
    to_UnPool(
        "up3",
        offset="(2,0,0)",
        to="(dwdeconv4-east)",
        height=H32,
        depth=D2048,
        width=W16,
        caption="",
    ),
    to_Conv(
        "dwdeconv5",
        s_filer="",
        n_filer=1,
        offset="(0,0,0)",
        to="(up3-east)",
        height=H32,
        depth=D2048,
        width=W1,
        caption="deconv3",
    ),
    to_Conv(
        "dwdeconv6",
        s_filer="{{2048×32}}",
        n_filer=8,
        offset="(0,0,0)",
        to="(dwdeconv5-east)",
        height=H32,
        depth=D2048,
        width=W8,
        caption="",
    ),
    to_Conv(
        "outconv",
        s_filer="{{2048×32}}",
        n_filer=1,
        offset="(2,0,0)",
        to="(dwdeconv6-east)",
        height=H32,
        depth=D2048,
        width=W1,
        caption="deconv4",
    ),
    # to_connection("up2", "deconv2"),
    # Output
    to_Conv(
        "out",
        s_filer="{{2048×32}}",
        n_filer=1,
        offset="(2,0,0)",
        to="(outconv-east)",
        height=H32,
        depth=D2048,
        width=W1,
        caption="output",
    ),
    # to_connection("deconv2", "out"),
    to_end(),
]


def main():
    name = "subter_lenet_arch"
    to_generate(arch, name + ".tex")


if __name__ == "__main__":
    main()