app.py

#!/usr/bin/env python

from __future__ import annotations

import os
import random
import shlex
import subprocess
import sys

import gradio as gr
import numpy as np
import torch
import torch.nn as nn
from huggingface_hub import hf_hub_download

if os.environ.get("SYSTEM") == "spaces":
    with open("patch") as f:
        subprocess.run(shlex.split("patch -p1"), cwd="stylegan2-pytorch", stdin=f)
    if not torch.cuda.is_available():
        with open("patch-cpu") as f:
            subprocess.run(shlex.split("patch -p1"), cwd="stylegan2-pytorch", stdin=f)

sys.path.insert(0, "stylegan2-pytorch")

from model import Generator

DESCRIPTION = """# [TADNE](https://thisanimedoesnotexist.ai/) (This Anime Does Not Exist) interpolation

Related Apps:
- [TADNE](https://huggingface.co/spaces/hysts/TADNE)
- [TADNE Image Viewer](https://huggingface.co/spaces/hysts/TADNE-image-viewer)
- [TADNE Image Selector](https://huggingface.co/spaces/hysts/TADNE-image-selector)
- [TADNE Image Search with DeepDanbooru](https://huggingface.co/spaces/hysts/TADNE-image-search-with-DeepDanbooru)
"""

MAX_SEED = np.iinfo(np.int32).max


def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    return seed


def load_model(device: torch.device) -> nn.Module:
    model = Generator(512, 1024, 4, channel_multiplier=2)
    path = hf_hub_download("public-data/TADNE", "models/aydao-anime-danbooru2019s-512-5268480.pt")
    checkpoint = torch.load(path)
    model.load_state_dict(checkpoint["g_ema"])
    model.eval()
    model.to(device)
    model.latent_avg = checkpoint["latent_avg"].to(device)
    with torch.inference_mode():
        z = torch.zeros((1, model.style_dim)).to(device)
        model([z], truncation=0.7, truncation_latent=model.latent_avg)
    return model


device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
model = load_model(device)


def generate_z(z_dim: int, seed: int) -> torch.Tensor:
    return torch.from_numpy(np.random.RandomState(seed).randn(1, z_dim)).float()


@torch.inference_mode()
def generate_image(z: torch.Tensor, truncation_psi: float, randomize_noise: bool) -> np.ndarray:
    out, _ = model([z], truncation=truncation_psi, truncation_latent=model.latent_avg, randomize_noise=randomize_noise)
    out = (out.permute(0, 2, 3, 1) * 127.5 + 128).clamp(0, 255).to(torch.uint8)
    return out[0].cpu().numpy()


@torch.inference_mode()
def generate_interpolated_images(
    seed0: int,
    seed1: int,
    num_intermediate: int,
    psi0: float,
    psi1: float,
    randomize_noise: bool,
) -> list[np.ndarray]:
    seed0 = int(np.clip(seed0, 0, MAX_SEED))
    seed1 = int(np.clip(seed1, 0, MAX_SEED))

    z0 = generate_z(model.style_dim, seed0)
    z1 = generate_z(model.style_dim, seed1)
    z0 = z0.to(device)
    z1 = z1.to(device)

    vec = z1 - z0
    dvec = vec / (num_intermediate + 1)
    zs = [z0 + dvec * i for i in range(num_intermediate + 2)]
    dpsi = (psi1 - psi0) / (num_intermediate + 1)
    psis = [psi0 + dpsi * i for i in range(num_intermediate + 2)]
    res = []
    for z, psi in zip(zs, psis):
        out = generate_image(z, psi, randomize_noise)
        res.append(out)
    return res


examples = [
    [29703, 55376, 3, 0.7, 0.7, False],
    [34141, 36864, 5, 0.7, 0.7, False],
    [74650, 88322, 7, 0.7, 0.7, False],
    [84314, 70317410, 9, 0.7, 0.7, False],
    [55376, 55376, 5, 0.3, 1.3, False],
]

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(DESCRIPTION)
    with gr.Row():
        with gr.Column():
            seed_1 = gr.Slider(label="Seed 1", minimum=0, maximum=MAX_SEED, step=1, value=29703)
            seed_2 = gr.Slider(label="Seed 2", minimum=0, maximum=MAX_SEED, step=1, value=55376)
            num_intermediate_frames = gr.Slider(
                label="Number of Intermediate Frames",
                minimum=1,
                maximum=21,
                step=1,
                value=3,
            )
            psi_1 = gr.Slider(label="Truncation psi 1", minimum=0, maximum=2, step=0.05, value=0.7)
            psi_2 = gr.Slider(label="Truncation psi 2", minimum=0, maximum=2, step=0.05, value=0.7)
            randomize_noise = gr.Checkbox(label="Randomize Noise", value=False)
            run_button = gr.Button("Run")
        with gr.Column():
            result = gr.Gallery(label="Output")

    inputs = [
        seed_1,
        seed_2,
        num_intermediate_frames,
        psi_1,
        psi_2,
        randomize_noise,
    ]
    gr.Examples(
        examples=examples,
        inputs=inputs,
        outputs=result,
        fn=generate_interpolated_images,
        cache_examples=os.getenv("CACHE_EXAMPLES") == "1",
    )
    run_button.click(
        fn=generate_interpolated_images,
        inputs=inputs,
        outputs=result,
        api_name="run",
    )

if __name__ == "__main__":
    demo.queue(max_size=10).launch()