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
| import gradio as gr from diffusers import DiffusionPipeline, FluxPipeline, AutoPipelineForText2Image, DPMSolverMultistepScheduler, HunyuanDiTPipeline from huggingface_hub import login import torch import uuid from pathlib import Path import time
login(token="hf_token")
MODEL_CONFIGS = { "Stable Diffusion v1-4": "CompVis/stable-diffusion-v1-4", "Stable Diffusion v1-5": "sd-legacy/stable-diffusion-v1-5", "Stable Diffusion v2-1": "stabilityai/stable-diffusion-2-1", "SDXL-Turbo": "stabilityai/sdxl-turbo", "Hunyuan-DiT": "Tencent-Hunyuan/HunyuanDiT-v1.1-Diffusers-Distilled", "FLUX.1": "black-forest-labs/FLUX.1-dev" }
output_dir = Path("output") output_dir.mkdir(exist_ok=True)
if torch.cuda.is_available(): torch_dtype = torch.float16 device = "cuda" torch.cuda.empty_cache() else: torch_dtype = torch.float32 device = "cpu"
def load_flux_model(): """Load FLUX.1 model""" try: print("Loading FLUX.1...") pipe = FluxPipeline.from_pretrained("black-forest-labs/FLUX.1-dev", torch_dtype=torch.bfloat16) print("FLUX.1 loaded successfully") return pipe except Exception as e: print(f"Failed to load FLUX.1: {str(e)}") return None
def load_sdxl_turbo_model(): """Load SDXL-Turbo model""" try: print("Loading SDXL-Turbo...") pipe = AutoPipelineForText2Image.from_pretrained( "stabilityai/sdxl-turbo", torch_dtype=torch_dtype, variant="fp16" ) pipe = pipe.to(device) print("SDXL-Turbo loaded successfully") return pipe except Exception as e: print(f"Failed to load SDXL-Turbo: {str(e)}") return None
def load_sd_v14_model(): """Load Stable Diffusion v1-4 model""" try: print("Loading Stable Diffusion v1-4...") pipe = DiffusionPipeline.from_pretrained( "CompVis/stable-diffusion-v1-4", torch_dtype=torch_dtype, ) pipe = pipe.to(device) print("Stable Diffusion v1-4 loaded successfully") return pipe except Exception as e: print(f"Failed to load Stable Diffusion v1-4: {str(e)}") return None
def load_sd_v15_model(): """Load Stable Diffusion v1-5 model""" try: print("Loading Stable Diffusion v1-5...") pipe = DiffusionPipeline.from_pretrained( "sd-legacy/stable-diffusion-v1-5", torch_dtype=torch_dtype, ) pipe = pipe.to(device) print("Stable Diffusion v1-5 loaded successfully") return pipe except Exception as e: print(f"Failed to load Stable Diffusion v1-5: {str(e)}") return None
def load_sd_v21_model(): """Load Stable Diffusion v2-1 model""" try: print("Loading Stable Diffusion v2-1...") pipe = DiffusionPipeline.from_pretrained( "stabilityai/stable-diffusion-2-1", torch_dtype=torch_dtype, ) pipe.scheduler = DPMSolverMultistepScheduler.from_config(pipe.scheduler.config) pipe = pipe.to(device) print("Stable Diffusion v2-1 loaded successfully") return pipe except Exception as e: print(f"Failed to load Stable Diffusion v2-1: {str(e)}") return None
def load_hunyuan_dit_model(): """Load Hunyuan-DiT model""" try: print("Loading Hunyuan-DiT...") pipe = HunyuanDiTPipeline.from_pretrained( "Tencent-Hunyuan/HunyuanDiT-v1.1-Diffusers-Distilled", torch_dtype=torch_dtype ) pipe = pipe.to(device) print("Hunyuan-DiT loaded successfully") return pipe except Exception as e: print(f"Failed to load Hunyuan-DiT: {str(e)}") return None
def load_model(model_name): """Load a model temporarily""" if "flux" in model_name.lower(): return load_flux_model() elif "turbo" in model_name.lower(): return load_sdxl_turbo_model() elif "2-1" in model_name.lower(): return load_sd_v21_model() elif "1-4" in model_name.lower(): return load_sd_v14_model() elif "1-5" in model_name.lower(): return load_sd_v15_model() elif "hunyuan" in model_name.lower(): return load_hunyuan_dit_model() else: print(f"Unknown model: {model_name}") return None
def unload_model(pipe): """Unload a model to free memory""" del pipe
if torch.cuda.is_available(): torch.cuda.empty_cache()
print("Model unloaded and memory cleared")
def generate_images_grid(prompt, selected_models=None): """Generate images for selected models and return individual outputs for grid layout""" sd14_img, sd15_img, sd21_img, turbo_img, hunyuan_img, flux_img = None, None, None, None, None, None sd14_time, sd15_time, sd21_time, turbo_time, hunyuan_time, flux_time = "", "", "", "", "", "" if not selected_models: return sd14_img, sd15_img, sd21_img, turbo_img, hunyuan_img, flux_img, sd14_time, sd15_time, sd21_time, turbo_time, hunyuan_time, flux_time
filename = f"{uuid.uuid4().hex}.png"
model_order = [ "Stable Diffusion v1-4", "Stable Diffusion v1-5", "Stable Diffusion v2-1", "SDXL-Turbo", "Hunyuan-DiT", "FLUX.1" ] print(f"Processing prompt: {prompt}") for model_name in model_order: if model_name not in selected_models: continue
pipe = None try: pipe = load_model(model_name) if pipe is None: continue
start_time = time.time() if "flux" in model_name.lower(): image = pipe( prompt, height=1024, width=1024, guidance_scale=3.5, num_inference_steps=50, max_sequence_length=512, generator=torch.Generator("cpu").manual_seed(0) ).images[0] elif "turbo" in model_name.lower(): image = pipe(prompt, height=1024, width=1024, num_inference_steps=1, guidance_scale=0.0).images[0] else: image = pipe(prompt, height=1024, width=1024).images[0]
end_time = time.time() duration = end_time - start_time
print(f"{model_name} generation time: {duration:.2f}s")
model_dir = output_dir / model_name.replace("/", "_").replace(".", "_") model_dir.mkdir(exist_ok=True)
image_path = model_dir / filename image.save(str(image_path))
if "flux" in model_name.lower(): flux_img = image flux_time = f"{duration:.2f}s" elif "turbo" in model_name.lower(): turbo_img = image turbo_time = f"{duration:.2f}s" elif "1-4" in model_name.lower(): sd14_img = image sd14_time = f"{duration:.2f}s" elif "1-5" in model_name.lower(): sd15_img = image sd15_time = f"{duration:.2f}s" elif "2-1" in model_name.lower(): sd21_img = image sd21_time = f"{duration:.2f}s" elif "hunyuan" in model_name.lower(): hunyuan_img = image hunyuan_time = f"{duration:.2f}s"
unload_model(pipe) pipe = None
except Exception as e: print(f"Error generating image with {model_name}: {str(e)}")
if pipe is not None: unload_model(pipe)
return sd14_img, sd15_img, sd21_img, turbo_img, hunyuan_img, flux_img, sd14_time, sd15_time, sd21_time, turbo_time, hunyuan_time, flux_time
example_prompts = [ "A pot of green plants grows in a red flower pot, centered and realistic.", "A lovely rabbit eating carrots. The photo is centered and in cartoon style.", "A green leaf is centered on a white background with clear texture", "A brown and white hamster standing in front of a white background. The photo is centered and in cartoon style." ]
with gr.Blocks(title="Text-to-Image Model Comparison") as demo: gr.Markdown("# Text-to-Image Model Comparison") gr.Markdown("Enter a prompt and generate images with different models. Images are saved in model-specific folders.")
with gr.Row(): prompt_input = gr.Textbox( label="Prompt", value=example_prompts[0], lines=3 )
with gr.Row(): model_selection = gr.CheckboxGroup( choices=list(MODEL_CONFIGS.keys()), value=["Stable Diffusion v1-4", "Stable Diffusion v1-5", "Stable Diffusion v2-1", "SDXL-Turbo", "Hunyuan-DiT"], label="Select Models" )
gr.Examples( examples=example_prompts, inputs=prompt_input, label="Example Prompts", examples_per_page=4 )
with gr.Row(): generate_btn = gr.Button("Generate Images", variant="primary")
with gr.Row(): with gr.Column(): gr.Markdown("### Stable Diffusion v1-4") sd14_image = gr.Image(label="Generated Image", height=512) sd14_timing = gr.Textbox(label="Generation Time", interactive=False) with gr.Column(): gr.Markdown("### Stable Diffusion v1-5") sd15_image = gr.Image(label="Generated Image", height=512) sd15_timing = gr.Textbox(label="Generation Time", interactive=False) with gr.Column(): gr.Markdown("### Stable Diffusion v2-1") sd21_image = gr.Image(label="Generated Image", height=512) sd21_timing = gr.Textbox(label="Generation Time", interactive=False) with gr.Row(): with gr.Column(): gr.Markdown("### SDXL-Turbo") turbo_image = gr.Image(label="Generated Image", height=512) turbo_timing = gr.Textbox(label="Generation Time", interactive=False) with gr.Column(): gr.Markdown("### Hunyuan-DiT") hunyuan_image = gr.Image(label="Generated Image", height=512) hunyuan_timing = gr.Textbox(label="Generation Time", interactive=False) with gr.Column(): gr.Markdown("### FLUX.1") flux_image = gr.Image(label="Generated Image", height=512) flux_timing = gr.Textbox(label="Generation Time", interactive=False)
generate_btn.click( fn=generate_images_grid, inputs=[prompt_input, model_selection], outputs=[sd14_image, sd15_image, sd21_image, turbo_image, hunyuan_image, flux_image, sd14_timing, sd15_timing, sd21_timing, turbo_timing, hunyuan_timing, flux_timing] )
if __name__ == "__main__": demo.launch(server_name="0.0.0.0")
|