v3p2

app.py

@@ -14,9 +14,206 @@ from datetime import datetime
 from diffusers.models import AutoencoderKL
 from diffusers import StableDiffusionXLPipeline, StableDiffusionXLImg2ImgPipeline
 
-# ... (keep all the imports and initial setup)
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+DESCRIPTION = "PonyDiffusion V6 XL"
+if not torch.cuda.is_available():
+    DESCRIPTION += "\n<p>Running on CPU 🥶 This demo does not work on CPU. </p>"
+IS_COLAB = utils.is_google_colab() or os.getenv("IS_COLAB") == "1"
+HF_TOKEN = os.getenv("HF_TOKEN")
+CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES") == "1"
+MIN_IMAGE_SIZE = int(os.getenv("MIN_IMAGE_SIZE", "512"))
+MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "2048"))
+USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE") == "1"
+ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD") == "1"
+OUTPUT_DIR = os.getenv("OUTPUT_DIR", "./outputs")
+
+MODEL = os.getenv(
+    "MODEL",
+    "https://huggingface.co/AstraliteHeart/pony-diffusion-v6/blob/main/v6.safetensors",
+)
+
+torch.backends.cudnn.deterministic = True
+torch.backends.cudnn.benchmark = False
+
+device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
+
+def load_pipeline(model_name):
+    vae = AutoencoderKL.from_pretrained(
+        "madebyollin/sdxl-vae-fp16-fix",
+        torch_dtype=torch.float16,
+    )
+    pipeline = (
+        StableDiffusionXLPipeline.from_single_file
+        if MODEL.endswith(".safetensors")
+        else StableDiffusionXLPipeline.from_pretrained
+    )
+
+    pipe = pipeline(
+        model_name,
+        vae=vae,
+        torch_dtype=torch.float16,
+        custom_pipeline="lpw_stable_diffusion_xl",
+        use_safetensors=True,
+        add_watermarker=False,
+        use_auth_token=HF_TOKEN,
+        variant="fp16",
+    )
+
+    pipe.to(device)
+    return pipe
+
+def parse_json_parameters(json_str):
+    try:
+        params = json.loads(json_str)
+        return params
+    except json.JSONDecodeError:
+        return None
+
+def apply_json_parameters(json_str):
+    params = parse_json_parameters(json_str)
+    if params:
+        return (
+            params.get("prompt", ""),
+            params.get("negative_prompt", ""),
+            params.get("seed", 0),
+            params.get("width", 1024),
+            params.get("height", 1024),
+            params.get("guidance_scale", 7.0),
+            params.get("num_inference_steps", 30),
+            params.get("sampler", "DPM++ 2M SDE Karras"),
+            params.get("aspect_ratio", "1024 x 1024"),
+            params.get("use_upscaler", False),
+            params.get("upscaler_strength", 0.55),
+            params.get("upscale_by", 1.5),
+        )
+    return [gr.update()] * 12
+
+def generate(
+    prompt: str,
+    negative_prompt: str = "",
+    seed: int = 0,
+    custom_width: int = 1024,
+    custom_height: int = 1024,
+    guidance_scale: float = 7.0,
+    num_inference_steps: int = 30,
+    sampler: str = "DPM++ 2M SDE Karras",
+    aspect_ratio_selector: str = "1024 x 1024",
+    use_upscaler: bool = False,
+    upscaler_strength: float = 0.55,
+    upscale_by: float = 1.5,
+    progress=gr.Progress(track_tqdm=True),
+) -> Image:
+    generator = utils.seed_everything(seed)
+
+    width, height = utils.aspect_ratio_handler(
+        aspect_ratio_selector,
+        custom_width,
+        custom_height,
+    )
+
+    width, height = utils.preprocess_image_dimensions(width, height)
+
+    backup_scheduler = pipe.scheduler
+    pipe.scheduler = utils.get_scheduler(pipe.scheduler.config, sampler)
+
+    if use_upscaler:
+        upscaler_pipe = StableDiffusionXLImg2ImgPipeline(**pipe.components)
+    metadata = {
+        "prompt": prompt,
+        "negative_prompt": negative_prompt,
+        "resolution": f"{width} x {height}",
+        "guidance_scale": guidance_scale,
+        "num_inference_steps": num_inference_steps,
+        "seed": seed,
+        "sampler": sampler,
+    }
+
+    if use_upscaler:
+        new_width = int(width * upscale_by)
+        new_height = int(height * upscale_by)
+        metadata["use_upscaler"] = {
+            "upscale_method": "nearest-exact",
+            "upscaler_strength": upscaler_strength,
+            "upscale_by": upscale_by,
+            "new_resolution": f"{new_width} x {new_height}",
+        }
+    else:
+        metadata["use_upscaler"] = None
+    logger.info(json.dumps(metadata, indent=4))
+
+    try:
+        if use_upscaler:
+            latents = pipe(
+                prompt=prompt,
+                negative_prompt=negative_prompt,
+                width=width,
+                height=height,
+                guidance_scale=guidance_scale,
+                num_inference_steps=num_inference_steps,
+                generator=generator,
+                output_type="latent",
+            ).images
+            upscaled_latents = utils.upscale(latents, "nearest-exact", upscale_by)
+            images = upscaler_pipe(
+                prompt=prompt,
+                negative_prompt=negative_prompt,
+                image=upscaled_latents,
+                guidance_scale=guidance_scale,
+                num_inference_steps=num_inference_steps,
+                strength=upscaler_strength,
+                generator=generator,
+                output_type="pil",
+            ).images
+        else:
+            images = pipe(
+                prompt=prompt,
+                negative_prompt=negative_prompt,
+                width=width,
+                height=height,
+                guidance_scale=guidance_scale,
+                num_inference_steps=num_inference_steps,
+                generator=generator,
+                output_type="pil",
+            ).images
+
+        if images and IS_COLAB:
+            for image in images:
+                filepath = utils.save_image(image, metadata, OUTPUT_DIR)
+                logger.info(f"Image saved as {filepath} with metadata")
+
+        # Update history after generation
+        history = gr.get_state("history") or []
+        history.insert(0, {"prompt": prompt, "image": images[0], "metadata": metadata})
+        gr.set_state("history", history[:10])  # Keep only the last 10 entries
+
+        return images, metadata, gr.update(choices=[h["prompt"] for h in history])
+    except Exception as e:
+        logger.exception(f"An error occurred: {e}")
+        raise
+    finally:
+        if use_upscaler:
+            del upscaler_pipe
+        pipe.scheduler = backup_scheduler
+        utils.free_memory()
+
+def get_random_prompt():
+    anime_characters = [
+        "Naruto Uzumaki", "Monkey D. Luffy", "Goku", "Eren Yeager", "Light Yagami",
+        "Lelouch Lamperouge", "Edward Elric", "Levi Ackerman", "Spike Spiegel",
+        "Sakura Haruno", "Mikasa Ackerman", "Asuka Langley Soryu", "Rem", "Megumin",
+        "Violet Evergarden"
+    ]
+    styles = ["pixel art", "stylized anime", "digital art", "watercolor", "sketch"]
+    scores = ["score_9", "score_8_up", "score_7_up"]
+    
+    character = random.choice(anime_characters)
+    style = random.choice(styles)
+    score = ", ".join(random.sample(scores, k=3))
+    
+    return f"{score}, {character}, {style}, show accurate"
 
-# ... (keep all the functions like load_pipeline, parse_json_parameters, apply_json_parameters, generate, get_random_prompt)
 
 if torch.cuda.is_available():
     pipe = load_pipeline(MODEL)