Spaces:
Running
on
Zero
Running
on
Zero
File size: 6,396 Bytes
fca8815 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 |
import argparse
from PIL import Image
import os
from src.flux.xflux_pipeline import XFluxPipeline
def create_argparser():
parser = argparse.ArgumentParser()
parser.add_argument(
"--prompt", type=str, required=True,
help="The input text prompt"
)
parser.add_argument(
"--neg_prompt", type=str, default="",
help="The input text negative prompt"
)
parser.add_argument(
"--img_prompt", type=str, default=None,
help="Path to input image prompt"
)
parser.add_argument(
"--neg_img_prompt", type=str, default=None,
help="Path to input negative image prompt"
)
parser.add_argument(
"--ip_scale", type=float, default=1.0,
help="Strength of input image prompt"
)
parser.add_argument(
"--neg_ip_scale", type=float, default=1.0,
help="Strength of negative input image prompt"
)
parser.add_argument(
"--local_path", type=str, default=None,
help="Local path to the model checkpoint (Controlnet)"
)
parser.add_argument(
"--repo_id", type=str, default=None,
help="A HuggingFace repo id to download model (Controlnet)"
)
parser.add_argument(
"--name", type=str, default=None,
help="A filename to download from HuggingFace"
)
parser.add_argument(
"--ip_repo_id", type=str, default=None,
help="A HuggingFace repo id to download model (IP-Adapter)"
)
parser.add_argument(
"--ip_name", type=str, default=None,
help="A IP-Adapter filename to download from HuggingFace"
)
parser.add_argument(
"--ip_local_path", type=str, default=None,
help="Local path to the model checkpoint (IP-Adapter)"
)
parser.add_argument(
"--lora_repo_id", type=str, default=None,
help="A HuggingFace repo id to download model (LoRA)"
)
parser.add_argument(
"--lora_name", type=str, default=None,
help="A LoRA filename to download from HuggingFace"
)
parser.add_argument(
"--lora_local_path", type=str, default=None,
help="Local path to the model checkpoint (Controlnet)"
)
parser.add_argument(
"--device", type=str, default="cuda",
help="Device to use (e.g. cpu, cuda:0, cuda:1, etc.)"
)
parser.add_argument(
"--offload", action='store_true', help="Offload model to CPU when not in use"
)
parser.add_argument(
"--use_ip", action='store_true', help="Load IP model"
)
parser.add_argument(
"--use_lora", action='store_true', help="Load Lora model"
)
parser.add_argument(
"--use_controlnet", action='store_true', help="Load Controlnet model"
)
parser.add_argument(
"--num_images_per_prompt", type=int, default=1,
help="The number of images to generate per prompt"
)
parser.add_argument(
"--image", type=str, default=None, help="Path to image"
)
parser.add_argument(
"--lora_weight", type=float, default=0.9, help="Lora model strength (from 0 to 1.0)"
)
parser.add_argument(
"--control_type", type=str, default="canny",
choices=("canny", "openpose", "depth", "hed", "hough", "tile"),
help="Name of controlnet condition, example: canny"
)
parser.add_argument(
"--model_type", type=str, default="flux-dev",
choices=("flux-dev", "flux-dev-fp8", "flux-schnell"),
help="Model type to use (flux-dev, flux-dev-fp8, flux-schnell)"
)
parser.add_argument(
"--width", type=int, default=1024, help="The width for generated image"
)
parser.add_argument(
"--height", type=int, default=1024, help="The height for generated image"
)
parser.add_argument(
"--num_steps", type=int, default=25, help="The num_steps for diffusion process"
)
parser.add_argument(
"--guidance", type=float, default=4, help="The guidance for diffusion process"
)
parser.add_argument(
"--seed", type=int, default=123456789, help="A seed for reproducible inference"
)
parser.add_argument(
"--true_gs", type=float, default=3.5, help="true guidance"
)
parser.add_argument(
"--timestep_to_start_cfg", type=int, default=5, help="timestep to start true guidance"
)
parser.add_argument(
"--save_path", type=str, default='results', help="Path to save"
)
return parser
def main(args):
if args.image:
image = Image.open(args.image)
else:
image = None
xflux_pipeline = XFluxPipeline(args.model_type, args.device, args.offload)
if args.use_ip:
print('load ip-adapter:', args.ip_local_path, args.ip_repo_id, args.ip_name)
xflux_pipeline.set_ip(args.ip_local_path, args.ip_repo_id, args.ip_name)
if args.use_lora:
print('load lora:', args.lora_local_path, args.lora_repo_id, args.lora_name)
xflux_pipeline.set_lora(args.lora_local_path, args.lora_repo_id, args.lora_name, args.lora_weight)
if args.use_controlnet:
print('load controlnet:', args.local_path, args.repo_id, args.name)
xflux_pipeline.set_controlnet(args.control_type, args.local_path, args.repo_id, args.name)
image_prompt = Image.open(args.img_prompt) if args.img_prompt else None
neg_image_prompt = Image.open(args.neg_img_prompt) if args.neg_img_prompt else None
for _ in range(args.num_images_per_prompt):
result = xflux_pipeline(
prompt=args.prompt,
controlnet_image=image,
width=args.width,
height=args.height,
guidance=args.guidance,
num_steps=args.num_steps,
seed=args.seed,
true_gs=args.true_gs,
neg_prompt=args.neg_prompt,
timestep_to_start_cfg=args.timestep_to_start_cfg,
image_prompt=image_prompt,
neg_image_prompt=neg_image_prompt,
ip_scale=args.ip_scale,
neg_ip_scale=args.neg_ip_scale,
)
if not os.path.exists(args.save_path):
os.mkdir(args.save_path)
ind = len(os.listdir(args.save_path))
result.save(os.path.join(args.save_path, f"result_{ind}.png"))
args.seed = args.seed + 1
if __name__ == "__main__":
args = create_argparser().parse_args()
main(args)
|