File size: 4,340 Bytes
04e327c | 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 | import argparse
import os
def get_token() -> str:
tok = os.environ.get("HF_TOKEN", "")
if tok:
return tok
for p in ("/content/king_hf_token.txt", "/content/hf_token.txt"):
if os.path.isfile(p):
return open(p).read().strip()
try:
from google.colab import userdata
return userdata.get("HF_TOKEN")
except Exception:
pass
raise RuntimeError("HF_TOKEN not found (env, /content/king_hf_token.txt or Colab secret 'HF_TOKEN')")
PROMPTS = [
"a futuristic royal palace at sunset, highly detailed, 8k, golden hour, epic composition",
"majestic arabian knight on horseback, desert landscape, cinematic lighting, photorealistic, 4k",
"a cozy modern indoor living room with warm lighting, ultra detailed interior photography",
"a serene outdoor garden path with blooming flowers and golden sunlight, highly detailed",
"cosmic king on a throne of stars, nebula background, majestic, epic fantasy, concept art",
"grand mosque at night, illuminated, reflective pool, stars, ultra detailed, architectural photography",
]
def main() -> None:
ap = argparse.ArgumentParser()
ap.add_argument("--new_lora", default="RASHID778/king2-image")
ap.add_argument("--old_repo", default="RASHID778/king2-image-dataset")
ap.add_argument("--old_file", default="king2-original/pytorch_lora_weights.safetensors")
ap.add_argument("--output_repo", default="RASHID778/king2-image")
ap.add_argument("--steps", type=int, default=30)
ap.add_argument("--guidance", type=float, default=7.5)
ap.add_argument("--size", type=int, default=768)
ap.add_argument("--prompts", default="")
args = ap.parse_args()
import torch
from diffusers import AutoencoderKL, DiffusionPipeline
from huggingface_hub import HfApi, hf_hub_download, login
from PIL import Image, ImageDraw
from safetensors.torch import load_file
login(token=get_token())
api = HfApi()
device = "cuda"
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = DiffusionPipeline.from_pretrained(
"stabilityai/stable-diffusion-xl-base-1.0",
vae=vae,
torch_dtype=torch.float16,
variant="fp16",
use_safetensors=True,
).to(device)
prompts = [p for p in (args.prompts.splitlines() + PROMPTS) if p.strip()][:8]
pipe.load_lora_weights(args.new_lora, adapter_name="new")
old_path = hf_hub_download(
args.old_repo,
args.old_file,
repo_type="dataset",
local_dir="/content/king2-old",
)
old_sd = load_file(old_path, device="cpu")
pipe.load_lora_weights(lora_state_dict=old_sd, adapter_name="old")
print(f"[lora] loaded new + old adapters ({len(prompts)} prompts)", flush=True)
neg = "blurry, low quality, distorted, ugly, bad anatomy, watermark, text"
os.makedirs("/content/verify", exist_ok=True)
for i, prompt in enumerate(prompts):
grid = None
for adapter in ("new", "old"):
pipe.set_adapters([adapter])
img = pipe(
prompt,
num_inference_steps=args.steps,
guidance_scale=args.guidance,
height=args.size,
width=args.size,
negative_prompt=neg,
).images[0]
canvas = Image.new("RGB", (img.width, img.height + 40), (255, 255, 255))
canvas.paste(img, (0, 0))
draw = ImageDraw.Draw(canvas)
draw.text((10, img.height + 10), f"{adapter.upper()}", fill=(0, 0, 0))
grid = canvas if grid is None else Image.fromarray(
__import__("numpy").concatenate(
[__import__("numpy").asarray(grid), __import__("numpy").asarray(canvas)],
axis=1,
)
)
out = f"/content/verify/verify_{i:02d}.png"
grid.save(out)
print(f"[gen] {i} -> {out} :: {prompt[:80]}", flush=True)
api.upload_folder(
repo_id=args.output_repo,
folder_path="/content/verify",
path_in_repo="examples_v2",
commit_message="king2-image V2 old-vs-new verification images",
)
print("[ok] uploaded examples_v2/ (new on top, old below)", flush=True)
if __name__ == "__main__":
main() |