app.py

import warnings
import argparse
import os
from PIL import Image
import numpy as np
import torch

import stylegan2
from stylegan2 import utils
from huggingface_hub import hf_hub_download
import gradio as gr
import re

from types import SimpleNamespace

#Edited from run_generator.py to return PIL images instead of saving them to the disk.
def generate_images(G, args):
    latent_size, label_size = G.latent_size, G.label_size
    device = torch.device(args.gpu[0] if args.gpu else 'cpu')
    if device.index is not None:
        torch.cuda.set_device(device.index)
    G.to(device)
    if args.truncation_psi != 1:
        G.set_truncation(truncation_psi=args.truncation_psi)
    if len(args.gpu) > 1:
        warnings.warn(
            'Noise can not be randomized based on the seed ' +
            'when using more than 1 GPU device. Noise will ' +
            'now be randomized from default random state.'
        )
        G.random_noise()
        G = torch.nn.DataParallel(G, device_ids=args.gpu)
    else:
        noise_reference = G.static_noise()

    def get_batch(seeds):
        latents = []
        labels = []
        if len(args.gpu) <= 1:
            noise_tensors = [[] for _ in noise_reference]
        for seed in seeds:
            rnd = np.random.default_rng(seed)
            latents.append(torch.from_numpy(rnd.standard_normal(latent_size)))
            if len(args.gpu) <= 1:
                for i, ref in enumerate(noise_reference):
                    noise_tensors[i].append(
                        torch.from_numpy(rnd.standard_normal(tuple([*ref.size()[1:]])))
                    )
            if label_size:
                labels.append(torch.tensor([rnd.integers(0, label_size)]))
        latents = torch.stack(latents, dim=0).to(
            device=device, dtype=torch.float32)
        if labels:
            labels = torch.cat(labels, dim=0).to(
                device=device, dtype=torch.int64)
        else:
            labels = None
        if len(args.gpu) <= 1:
            noise_tensors = [
                torch.stack(noise, dim=0).to(
                    device=device, dtype=torch.float32)
                for noise in noise_tensors
            ]
        else:
            noise_tensors = None
        return latents, labels, noise_tensors
    return_images = []
    for i in range(0, len(args.seeds), args.batch_size):
        latents, labels, noise_tensors = get_batch(
            args.seeds[i: i + args.batch_size])
        if noise_tensors is not None:
            G.static_noise(noise_tensors=noise_tensors)
        with torch.no_grad():
            generated = G(latents, labels=labels)
        images = utils.tensor_to_PIL(
            generated, pixel_min=args.pixel_min, pixel_max=args.pixel_max)
        return_images.extend(images)
    return return_images
 
 
 #----------------------------------------------------------------------------
interface_modelversion_labels = [
    "TWDNEv3 iteration 24664 (best and current version on TWDNE)",
    "TWDNEv3 iteration 18528 (the most used version on the Internet)",
    "TWDNEv3 iteration 17325"
]

def inference(seed, truncation_psi, modelversion_label):
    model_iteration = re.search("TWDNEv3 iteration (\d{5})", modelversion_label).group(1)
    G = stylegan2.models.load(
        hf_hub_download("hr16/Gwern-TWDNEv3-pytorch_ckpt", f"iteration-{model_iteration}/Gs.pth", use_auth_token=os.environ['MODEL_READING_TOKEN'])
    )
    G.eval()
    return generate_images(
        G, 
        SimpleNamespace(**{
            'truncation_psi': truncation_psi,
            'seeds': [seed],
            'batch_size': 1,
            'pixel_min': -1,
            'pixel_max': 1,
            'gpu': []
        }) #Replace ArgumentParser at run_generator.py
    )[0]
    
title = "TWDNEv3 CPU Generator"
description = "Gradio Demo for TWDNEv3 CPU Generator (stylegan2_pytorch port)"
article = ""
gr.Interface(
    inference, 
    [
        gr.Number(precision=0, label="PCG64 PRNG Seed (any-bit-size unsigned int, note that it may different from the original site)"), 
        gr.Slider(0, 2, step=0.1, value=0.7, label='Truncation psi (aka creative level, between 0 and 2)'),
        gr.Radio(
            interface_modelversion_labels,
            value="TWDNEv3 iteration 24664 (best and current version on TWDNE)", 
            type="value", 
            label="Model versions"         
        )
    ], 
    gr.outputs.Image(type="pil"),
    title=title,description=description,article=article,allow_flagging=False,allow_screenshot=False
).launch()