File size: 13,443 Bytes
338d95d |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 |
#!/usr/bin/env python3
"""
CompI Phase 1.E: Personal Style Generation with LoRA
Generate images using your trained LoRA personal style weights.
Usage:
python src/generators/compi_phase1e_style_generation.py --lora-path lora_models/my_style/checkpoint-1000
python src/generators/compi_phase1e_style_generation.py --help
"""
import os
import argparse
import json
from datetime import datetime
from pathlib import Path
from typing import Optional, List
import torch
from PIL import Image
from diffusers import StableDiffusionPipeline, DPMSolverMultistepScheduler
from peft import PeftModel
# -------- 1. CONFIGURATION --------
DEFAULT_MODEL = "runwayml/stable-diffusion-v1-5"
DEFAULT_STEPS = 30
DEFAULT_GUIDANCE = 7.5
DEFAULT_WIDTH = 512
DEFAULT_HEIGHT = 512
OUTPUT_DIR = "outputs"
# -------- 2. UTILITY FUNCTIONS --------
def setup_args():
"""Setup command line arguments."""
parser = argparse.ArgumentParser(
description="CompI Phase 1.E: Personal Style Generation with LoRA",
formatter_class=argparse.RawDescriptionHelpFormatter,
epilog="""
Examples:
# Generate with trained LoRA style
python %(prog)s --lora-path lora_models/my_style/checkpoint-1000 "a cat in my_style"
# Interactive mode
python %(prog)s --lora-path lora_models/my_style/checkpoint-1000 --interactive
# Multiple variations
python %(prog)s --lora-path lora_models/my_style/checkpoint-1000 "landscape" --variations 4
"""
)
parser.add_argument("prompt", nargs="*", help="Text prompt for generation")
parser.add_argument("--lora-path", required=True,
help="Path to trained LoRA checkpoint directory")
parser.add_argument("--model-name", default=DEFAULT_MODEL,
help=f"Base Stable Diffusion model (default: {DEFAULT_MODEL})")
parser.add_argument("--variations", "-v", type=int, default=1,
help="Number of variations to generate")
parser.add_argument("--steps", type=int, default=DEFAULT_STEPS,
help=f"Number of inference steps (default: {DEFAULT_STEPS})")
parser.add_argument("--guidance", type=float, default=DEFAULT_GUIDANCE,
help=f"Guidance scale (default: {DEFAULT_GUIDANCE})")
parser.add_argument("--width", type=int, default=DEFAULT_WIDTH,
help=f"Image width (default: {DEFAULT_WIDTH})")
parser.add_argument("--height", type=int, default=DEFAULT_HEIGHT,
help=f"Image height (default: {DEFAULT_HEIGHT})")
parser.add_argument("--seed", type=int,
help="Random seed for reproducible generation")
parser.add_argument("--negative", "-n", default="",
help="Negative prompt")
parser.add_argument("--lora-scale", type=float, default=1.0,
help="LoRA scale factor (0.0-2.0, default: 1.0)")
parser.add_argument("--interactive", "-i", action="store_true",
help="Interactive mode")
parser.add_argument("--output-dir", default=OUTPUT_DIR,
help=f"Output directory (default: {OUTPUT_DIR})")
parser.add_argument("--list-styles", action="store_true",
help="List available LoRA styles")
return parser.parse_args()
def load_lora_info(lora_path: str) -> dict:
"""Load LoRA training information."""
lora_dir = Path(lora_path)
# Try to find training info
info_files = [
lora_dir / "training_info.json",
lora_dir.parent / "training_info.json"
]
for info_file in info_files:
if info_file.exists():
with open(info_file) as f:
return json.load(f)
# Fallback info
return {
'style_name': lora_dir.parent.name,
'model_name': DEFAULT_MODEL,
'lora_rank': 4,
'lora_alpha': 32
}
def load_pipeline_with_lora(model_name: str, lora_path: str, device: str):
"""Load Stable Diffusion pipeline with LoRA weights."""
print(f"π Loading base model: {model_name}")
# Load base pipeline
pipe = StableDiffusionPipeline.from_pretrained(
model_name,
torch_dtype=torch.float16 if device == "cuda" else torch.float32,
safety_checker=None,
requires_safety_checker=False
)
# Use DPM solver for faster inference
pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config)
print(f"π¨ Loading LoRA weights from: {lora_path}")
# Load LoRA weights
lora_dir = Path(lora_path)
if not lora_dir.exists():
raise FileNotFoundError(f"LoRA path not found: {lora_path}")
# Apply LoRA to UNet
pipe.unet = PeftModel.from_pretrained(pipe.unet, lora_path)
# Move to device
pipe = pipe.to(device)
# Enable memory efficient attention if available
if hasattr(pipe, "enable_xformers_memory_efficient_attention"):
try:
pipe.enable_xformers_memory_efficient_attention()
except Exception:
pass
return pipe
def generate_with_style(
pipe,
prompt: str,
negative_prompt: str = "",
num_inference_steps: int = DEFAULT_STEPS,
guidance_scale: float = DEFAULT_GUIDANCE,
width: int = DEFAULT_WIDTH,
height: int = DEFAULT_HEIGHT,
seed: Optional[int] = None,
lora_scale: float = 1.0
):
"""Generate image with LoRA style."""
# Set LoRA scale
if hasattr(pipe.unet, 'set_adapter_scale'):
pipe.unet.set_adapter_scale(lora_scale)
# Setup generator
if seed is not None:
generator = torch.Generator(device=pipe.device).manual_seed(seed)
else:
generator = None
seed = torch.seed()
# Generate image
with torch.autocast(pipe.device.type):
result = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
num_inference_steps=num_inference_steps,
guidance_scale=guidance_scale,
width=width,
height=height,
generator=generator
)
return result.images[0], seed
def save_generated_image(
image: Image.Image,
prompt: str,
style_name: str,
seed: int,
variation: int,
output_dir: str,
metadata: dict = None
):
"""Save generated image with metadata."""
# Create output directory
os.makedirs(output_dir, exist_ok=True)
# Generate filename
prompt_slug = "_".join(prompt.lower().split()[:5])
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
filename = f"{prompt_slug[:25]}_lora_{style_name}_{timestamp}_seed{seed}_v{variation}.png"
filepath = os.path.join(output_dir, filename)
# Save image
image.save(filepath)
# Save metadata if provided
if metadata:
metadata_file = filepath.replace('.png', '_metadata.json')
with open(metadata_file, 'w') as f:
json.dump(metadata, f, indent=2)
return filepath
def list_available_styles():
"""List available LoRA styles."""
lora_dir = Path("lora_models")
if not lora_dir.exists():
print("β No LoRA models directory found")
return
print("π¨ Available LoRA Styles:")
print("=" * 40)
styles_found = False
for style_dir in lora_dir.iterdir():
if style_dir.is_dir():
# Look for checkpoints
checkpoints = list(style_dir.glob("checkpoint-*"))
if checkpoints:
styles_found = True
latest_checkpoint = max(checkpoints, key=lambda x: int(x.name.split('-')[1]))
# Load info if available
info_file = style_dir / "training_info.json"
if info_file.exists():
with open(info_file) as f:
info = json.load(f)
print(f"π {style_dir.name}")
print(f" Latest: {latest_checkpoint.name}")
print(f" Steps: {info.get('total_steps', 'unknown')}")
print(f" Model: {info.get('model_name', 'unknown')}")
else:
print(f"π {style_dir.name}")
print(f" Latest: {latest_checkpoint.name}")
print()
if not styles_found:
print("β No trained LoRA styles found")
print("π‘ Train a style first using: python src/generators/compi_phase1e_lora_training.py")
def interactive_generation(pipe, lora_info: dict, args):
"""Interactive generation mode."""
style_name = lora_info.get('style_name', 'custom')
print(f"π¨ Interactive LoRA Style Generation - {style_name}")
print("=" * 50)
print("π‘ Tips:")
print(f" - Include '{style_name}' or trigger words in your prompts")
print(f" - Adjust LoRA scale (0.0-2.0) to control style strength")
print(" - Type 'quit' to exit")
print()
while True:
try:
# Get prompt
prompt = input("Enter prompt: ").strip()
if not prompt or prompt.lower() == 'quit':
break
# Get optional parameters
variations = input(f"Variations (default: 1): ").strip()
variations = int(variations) if variations.isdigit() else 1
lora_scale = input(f"LoRA scale (default: {args.lora_scale}): ").strip()
lora_scale = float(lora_scale) if lora_scale else args.lora_scale
# Generate images
print(f"π¨ Generating {variations} variation(s)...")
for i in range(variations):
image, seed = generate_with_style(
pipe, prompt, args.negative,
args.steps, args.guidance,
args.width, args.height,
args.seed, lora_scale
)
# Save image
filepath = save_generated_image(
image, prompt, style_name, seed, i + 1, args.output_dir,
{
'prompt': prompt,
'negative_prompt': args.negative,
'style_name': style_name,
'lora_scale': lora_scale,
'seed': seed,
'steps': args.steps,
'guidance_scale': args.guidance,
'timestamp': datetime.now().isoformat()
}
)
print(f"β
Saved: {filepath}")
print()
except KeyboardInterrupt:
break
except Exception as e:
print(f"β Error: {e}")
print()
def main():
"""Main function."""
args = setup_args()
# List styles if requested
if args.list_styles:
list_available_styles()
return 0
# Check LoRA path
if not os.path.exists(args.lora_path):
print(f"β LoRA path not found: {args.lora_path}")
return 1
# Load LoRA info
lora_info = load_lora_info(args.lora_path)
style_name = lora_info.get('style_name', 'custom')
print(f"π¨ CompI Phase 1.E: Personal Style Generation")
print(f"Style: {style_name}")
print("=" * 50)
# Setup device
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"π₯οΈ Using device: {device}")
# Load pipeline
try:
pipe = load_pipeline_with_lora(args.model_name, args.lora_path, device)
print("β
Pipeline loaded successfully")
except Exception as e:
print(f"β Failed to load pipeline: {e}")
return 1
# Interactive mode
if args.interactive:
interactive_generation(pipe, lora_info, args)
return 0
# Command line mode
prompt = " ".join(args.prompt) if args.prompt else input("Enter prompt: ").strip()
if not prompt:
print("β No prompt provided")
return 1
print(f"π¨ Generating {args.variations} variation(s) for: {prompt}")
# Generate images
for i in range(args.variations):
try:
image, seed = generate_with_style(
pipe, prompt, args.negative,
args.steps, args.guidance,
args.width, args.height,
args.seed, args.lora_scale
)
# Save image
filepath = save_generated_image(
image, prompt, style_name, seed, i + 1, args.output_dir,
{
'prompt': prompt,
'negative_prompt': args.negative,
'style_name': style_name,
'lora_scale': args.lora_scale,
'seed': seed,
'steps': args.steps,
'guidance_scale': args.guidance,
'timestamp': datetime.now().isoformat()
}
)
print(f"β
Generated variation {i + 1}: {filepath}")
except Exception as e:
print(f"β Error generating variation {i + 1}: {e}")
print("π Generation complete!")
return 0
if __name__ == "__main__":
exit(main())
|