ColossalAI/examples/images/diffusion/scripts/img2img.py

"""make variations of input image"""

import argparse
import os
from contextlib import nullcontext
from itertools import islice

import numpy as np
import PIL
import torch
from einops import rearrange, repeat
from omegaconf import OmegaConf
from PIL import Image
from torch import autocast
from torchvision.utils import make_grid
from tqdm import tqdm, trange

try:
    from lightning.pytorch import seed_everything
except:
    from pytorch_lightning import seed_everything

from imwatermark import WatermarkEncoder
from ldm.models.diffusion.ddim import DDIMSampler
from ldm.util import instantiate_from_config
from scripts.txt2img import put_watermark
from utils import replace_module


def chunk(it, size):
    it = iter(it)
    return iter(lambda: tuple(islice(it, size)), ())


def load_model_from_config(config, ckpt, verbose=False):
    print(f"Loading model from {ckpt}")
    pl_sd = torch.load(ckpt, map_location="cpu")
    if "global_step" in pl_sd:
        print(f"Global Step: {pl_sd['global_step']}")
    sd = pl_sd["state_dict"]
    model = instantiate_from_config(config.model)
    m, u = model.load_state_dict(sd, strict=False)
    if len(m) > 0 and verbose:
        print("missing keys:")
        print(m)
    if len(u) > 0 and verbose:
        print("unexpected keys:")
        print(u)

    model.eval()
    return model


def load_img(path):
    image = Image.open(path).convert("RGB")
    w, h = image.size
    print(f"loaded input image of size ({w}, {h}) from {path}")
    w, h = map(lambda x: x - x % 64, (w, h))  # resize to integer multiple of 64
    image = image.resize((w, h), resample=PIL.Image.LANCZOS)
    image = np.array(image).astype(np.float32) / 255.0
    image = image[None].transpose(0, 3, 1, 2)
    image = torch.from_numpy(image)
    return 2.0 * image - 1.0


def main():
    parser = argparse.ArgumentParser()

    parser.add_argument(
        "--prompt",
        type=str,
        nargs="?",
        default="a painting of a virus monster playing guitar",
        help="the prompt to render",
    )

    parser.add_argument("--init-img", type=str, nargs="?", help="path to the input image")

    parser.add_argument(
        "--outdir", type=str, nargs="?", help="dir to write results to", default="outputs/img2img-samples"
    )

    parser.add_argument(
        "--ddim_steps",
        type=int,
        default=50,
        help="number of ddim sampling steps",
    )

    parser.add_argument(
        "--fixed_code",
        action="store_true",
        help="if enabled, uses the same starting code across all samples ",
    )

    parser.add_argument(
        "--ddim_eta",
        type=float,
        default=0.0,
        help="ddim eta (eta=0.0 corresponds to deterministic sampling",
    )
    parser.add_argument(
        "--n_iter",
        type=int,
        default=1,
        help="sample this often",
    )

    parser.add_argument(
        "--C",
        type=int,
        default=4,
        help="latent channels",
    )
    parser.add_argument(
        "--f",
        type=int,
        default=8,
        help="downsampling factor, most often 8 or 16",
    )

    parser.add_argument(
        "--n_samples",
        type=int,
        default=2,
        help="how many samples to produce for each given prompt. A.k.a batch size",
    )

    parser.add_argument(
        "--n_rows",
        type=int,
        default=0,
        help="rows in the grid (default: n_samples)",
    )

    parser.add_argument(
        "--scale",
        type=float,
        default=9.0,
        help="unconditional guidance scale: eps = eps(x, empty) + scale * (eps(x, cond) - eps(x, empty))",
    )

    parser.add_argument(
        "--strength",
        type=float,
        default=0.8,
        help="strength for noising/unnoising. 1.0 corresponds to full destruction of information in init image",
    )

    parser.add_argument(
        "--from-file",
        type=str,
        help="if specified, load prompts from this file",
    )
    parser.add_argument(
        "--config",
        type=str,
        default="configs/stable-diffusion/v2-inference.yaml",
        help="path to config which constructs model",
    )
    parser.add_argument(
        "--ckpt",
        type=str,
        help="path to checkpoint of model",
    )
    parser.add_argument(
        "--seed",
        type=int,
        default=42,
        help="the seed (for reproducible sampling)",
    )
    parser.add_argument(
        "--precision", type=str, help="evaluate at this precision", choices=["full", "autocast"], default="autocast"
    )
    parser.add_argument(
        "--use_int8",
        type=bool,
        default=False,
        help="use int8 for inference",
    )

    opt = parser.parse_args()
    seed_everything(opt.seed)

    config = OmegaConf.load(f"{opt.config}")
    model = load_model_from_config(config, f"{opt.ckpt}")

    device = torch.device("cuda") if torch.cuda.is_available() else torch.device("cpu")
    model = model.to(device)

    # quantize model
    if opt.use_int8:
        model = replace_module(model)
        # # to compute the model size
        # getModelSize(model)

    sampler = DDIMSampler(model)

    os.makedirs(opt.outdir, exist_ok=True)
    outpath = opt.outdir

    print("Creating invisible watermark encoder (see https://github.com/ShieldMnt/invisible-watermark)...")
    wm = "SDV2"
    wm_encoder = WatermarkEncoder()
    wm_encoder.set_watermark("bytes", wm.encode("utf-8"))

    batch_size = opt.n_samples
    n_rows = opt.n_rows if opt.n_rows > 0 else batch_size
    if not opt.from_file:
        prompt = opt.prompt
        assert prompt is not None
        data = [batch_size * [prompt]]

    else:
        print(f"reading prompts from {opt.from_file}")
        with open(opt.from_file, "r") as f:
            data = f.read().splitlines()
            data = list(chunk(data, batch_size))

    sample_path = os.path.join(outpath, "samples")
    os.makedirs(sample_path, exist_ok=True)
    base_count = len(os.listdir(sample_path))
    grid_count = len(os.listdir(outpath)) - 1

    assert os.path.isfile(opt.init_img)
    init_image = load_img(opt.init_img).to(device)
    init_image = repeat(init_image, "1 ... -> b ...", b=batch_size)
    init_latent = model.get_first_stage_encoding(model.encode_first_stage(init_image))  # move to latent space

    sampler.make_schedule(ddim_num_steps=opt.ddim_steps, ddim_eta=opt.ddim_eta, verbose=False)

    assert 0.0 <= opt.strength <= 1.0, "can only work with strength in [0.0, 1.0]"
    t_enc = int(opt.strength * opt.ddim_steps)
    print(f"target t_enc is {t_enc} steps")

    precision_scope = autocast if opt.precision == "autocast" else nullcontext
    with torch.no_grad():
        with precision_scope("cuda"):
            with model.ema_scope():
                all_samples = list()
                for n in trange(opt.n_iter, desc="Sampling"):
                    for prompts in tqdm(data, desc="data"):
                        uc = None
                        if opt.scale != 1.0:
                            uc = model.get_learned_conditioning(batch_size * [""])
                        if isinstance(prompts, tuple):
                            prompts = list(prompts)
                        c = model.get_learned_conditioning(prompts)

                        # encode (scaled latent)
                        z_enc = sampler.stochastic_encode(init_latent, torch.tensor([t_enc] * batch_size).to(device))
                        # decode it
                        samples = sampler.decode(
                            z_enc,
                            c,
                            t_enc,
                            unconditional_guidance_scale=opt.scale,
                            unconditional_conditioning=uc,
                        )

                        x_samples = model.decode_first_stage(samples)
                        x_samples = torch.clamp((x_samples + 1.0) / 2.0, min=0.0, max=1.0)

                        for x_sample in x_samples:
                            x_sample = 255.0 * rearrange(x_sample.cpu().numpy(), "c h w -> h w c")
                            img = Image.fromarray(x_sample.astype(np.uint8))
                            img = put_watermark(img, wm_encoder)
                            img.save(os.path.join(sample_path, f"{base_count:05}.png"))
                            base_count += 1
                        all_samples.append(x_samples)

                # additionally, save as grid
                grid = torch.stack(all_samples, 0)
                grid = rearrange(grid, "n b c h w -> (n b) c h w")
                grid = make_grid(grid, nrow=n_rows)

                # to image
                grid = 255.0 * rearrange(grid, "c h w -> h w c").cpu().numpy()
                grid = Image.fromarray(grid.astype(np.uint8))
                grid = put_watermark(grid, wm_encoder)
                grid.save(os.path.join(outpath, f"grid-{grid_count:04}.png"))
                grid_count += 1

    print(f"Your samples are ready and waiting for you here: \n{outpath} \nEnjoy.")


if __name__ == "__main__":
    main()
    # # to compute the mem allocated
    # print(torch.cuda.max_memory_allocated() / 1024 / 1024)