DiffQRCoder/run_diffqrcoder.py at main · jwliao1209/DiffQRCoder · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
import os
from pathlib import Path
from argparse import ArgumentParser, Namespace

import torch
from diffusers import ControlNetModel, DDIMScheduler
from diffusers.utils import load_image

from diffqrcoder import DiffQRCoderPipeline


def parse_arguments() -> Namespace:
    parser = ArgumentParser()
    parser.add_argument(
        "--controlnet_ckpt",
        type=str,
        default="checkpoints/control_v1p_sd15_qrcode_monster"
    )
    parser.add_argument(
        "--pipe_ckpt",
        type=str,
        default="https://huggingface.co/fp16-guy/Cetus-Mix_Whalefall_fp16_cleaned/blob/main/cetusMix_Whalefall2_fp16.safetensors"
    )
    parser.add_argument(
        "--qrcode_path",
        type=str,
        default="qrcodes/thanks_reviewer.png"
    )
    parser.add_argument(
        "--qrcode_module_size",
        type=int,
        default=20,
    )
    parser.add_argument(
        "--qrcode_padding",
        type=int,
        default=78,
    )
    parser.add_argument(
        "--num_inference_steps",
        type=int,
        default=40,
    )
    parser.add_argument(
        "--prompt",
        type=str,
        default="Winter wonderland, fresh snowfall, evergreen trees, cozy log cabin, smoke rising from chimney, aurora borealis in night sky.",
    )
    parser.add_argument(
        "--neg_prompt",
        type=str,
        default="easynegative"
    )
    parser.add_argument(
        "--controlnet_conditioning_scale",
        type=float,
        default=1.35,
    )
    parser.add_argument(
        "-srg",
        "--scanning_robust_guidance_scale",
        type=float,
        default=500,
    )
    parser.add_argument(
        "-pg",
        "--perceptual_guidance_scale",
        type=float,
        default=2,
    )
    parser.add_argument(
        "--srmpgd_num_iteration",
        type=int,
        default=None,
    )
    parser.add_argument(
        "--srmpgd_lr",
        type=float,
        default=0.1,
    )
    parser.add_argument(
        "--device",
        type=str,
        default="cuda"
    )
    parser.add_argument(
        "--output_path",
        type=str,
        default="results/pycon.png",
    )
    return parser.parse_args()


if __name__ == "__main__":
    args = parse_arguments()
    os.makedirs(os.path.dirname(args.output_path), exist_ok=True)

    qrcode = load_image(args.qrcode_path)
    controlnet = ControlNetModel.from_pretrained(
        args.controlnet_ckpt,
        torch_dtype=torch.float16,
    )
    pipe = DiffQRCoderPipeline.from_single_file(
        args.pipe_ckpt,
        controlnet=controlnet,
        torch_dtype=torch.float16,
        use_auth_token=True,
    )
    pipe.scheduler = DDIMScheduler.from_config(pipe.scheduler.config)
    pipe = pipe.to(args.device)

    result = pipe(
        prompt=args.prompt,
        qrcode=qrcode,
        qrcode_module_size=args.qrcode_module_size,
        qrcode_padding=args.qrcode_padding,
        negative_prompt=args.neg_prompt,
        num_inference_steps=args.num_inference_steps,
        generator=torch.Generator(device=args.device).manual_seed(1),
        controlnet_conditioning_scale=args.controlnet_conditioning_scale,
        scanning_robust_guidance_scale=args.scanning_robust_guidance_scale,
        perceptual_guidance_scale=args.perceptual_guidance_scale,
        srmpgd_num_iteration=args.srmpgd_num_iteration,
        srmpgd_lr=args.srmpgd_lr,
    )
    result.images[0].save(args.output_path)