import os import random import uuid import json import gradio as gr import numpy as np from PIL import Image import spaces import torch from diffusers import StableDiffusionXLPipeline, EulerAncestralDiscreteScheduler from gradio.themes.base import Base from gradio.themes.utils import colors, fonts, sizes from typing import Iterable class Seafoam(Base): def __init__( self, *, primary_hue: colors.Color | str = colors.emerald, secondary_hue: colors.Color | str = colors.blue, neutral_hue: colors.Color | str = colors.gray, spacing_size: sizes.Size | str = sizes.spacing_md, radius_size: sizes.Size | str = sizes.radius_md, text_size: sizes.Size | str = sizes.text_lg, font: fonts.Font | str | Iterable[fonts.Font | str] = ( fonts.GoogleFont("Quicksand"), "ui-sans-serif", "sans-serif", ), font_mono: fonts.Font | str | Iterable[fonts.Font | str] = ( fonts.GoogleFont("IBM Plex Mono"), "ui-monospace", "monospace", ), ): super().__init__( primary_hue=primary_hue, secondary_hue=secondary_hue, neutral_hue=neutral_hue, spacing_size=spacing_size, radius_size=radius_size, text_size=text_size, font=font, font_mono=font_mono, ) seafoam = Seafoam() css = ''' .gradio-container { max-width: 100%; margin: 0 auto; } h1 { text-align: center; } footer { visibility: hidden; } ''' examples = [ "3d image, cute girl, in the style of Pixar --ar 1:2 --stylize 750, 4K resolution highlights, Sharp focus, octane render, ray tracing, Ultra-High-Definition, 8k, UHD, HDR, (Masterpiece:1.5), (best quality:1.5)", "A glass cup of cold coffee placed on a rustic wooden table, surrounded by soft morning light. The coffee is rich, dark, and topped with a light layer of creamy froth, droplets of condensation sliding down the glass.", "Vector illustration of a horse, vector graphic design with flat colors on an brown background in the style of vector art, using simple shapes and graphics with simple details, professionally designed as a tshirt logo ready for print on a white background. --ar 89:82 --v 6.0 --style raw", "Man in brown leather jacket posing for camera, in the style of sleek and stylized, clockpunk, subtle shades, exacting precision, ferrania p30 --ar 67:101 --v 5", "Commercial photography, giant burger, white lighting, studio light, 8k octane rendering, high resolution photography, insanely detailed, fine details, on white isolated plain, 8k, commercial photography, stock photo, professional color grading, --v 4 --ar 9:16 " ] MODEL_ID = os.getenv("MODEL_VAL_PATH") # SG161222/RealVisXL_V5.0_Lightning or SG161222/RealVisXL_V4.0_Lightning MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "4096")) USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE", "0") == "1" ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD", "0") == "1" BATCH_SIZE = int(os.getenv("BATCH_SIZE", "1")) device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") pipe = StableDiffusionXLPipeline.from_pretrained( MODEL_ID, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32, use_safetensors=True, add_watermarker=False, ).to(device) pipe.scheduler = EulerAncestralDiscreteScheduler.from_config(pipe.scheduler.config) if USE_TORCH_COMPILE: pipe.compile() if ENABLE_CPU_OFFLOAD: pipe.enable_model_cpu_offload() MAX_SEED = np.iinfo(np.int32).max def save_image(img): unique_name = str(uuid.uuid4()) + ".png" img.save(unique_name) return unique_name def randomize_seed_fn(seed: int, randomize_seed: bool) -> int: if randomize_seed: seed = random.randint(0, MAX_SEED) return seed @spaces.GPU def generate( prompt: str, negative_prompt: str = "", use_negative_prompt: bool = False, seed: int = 1, width: int = 1024, height: int = 1024, guidance_scale: float = 3, num_inference_steps: int = 25, randomize_seed: bool = False, use_resolution_binning: bool = True, num_images: int = 4, progress=gr.Progress(track_tqdm=True), ): seed = int(randomize_seed_fn(seed, randomize_seed)) generator = torch.Generator(device=device).manual_seed(seed) options = { "prompt": [prompt] * num_images, "negative_prompt": [negative_prompt] * num_images if use_negative_prompt else None, "width": width, "height": height, "guidance_scale": guidance_scale, "num_inference_steps": num_inference_steps, "generator": generator, "output_type": "pil", } if use_resolution_binning: options["use_resolution_binning"] = True images = [] for i in range(0, num_images, BATCH_SIZE): batch_options = options.copy() batch_options["prompt"] = options["prompt"][i:i+BATCH_SIZE] if "negative_prompt" in batch_options: batch_options["negative_prompt"] = options["negative_prompt"][i:i+BATCH_SIZE] images.extend(pipe(**batch_options).images) image_paths = [save_image(img) for img in images] return image_paths, seed with gr.Blocks(theme=seafoam, css=css) as demo: gr.Markdown("## STABLE HAMSTER") with gr.Row(): with gr.Column(scale=1): with gr.Row(): prompt = gr.Text( show_label=False, max_lines=1, placeholder="Enter your prompt", container=False, ) run_button = gr.Button("Run", scale=0, variant="primary") result = gr.Gallery(show_label=False, format="png", columns=2, object_fit="contain") with gr.Accordion("Advanced Settings", open=False): num_images = gr.Slider( label="Number of Images", minimum=1, maximum=4, step=1, value=4, ) with gr.Row(): use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=True) negative_prompt = gr.Text( label="Negative prompt", max_lines=5, lines=4, placeholder="Enter a negative prompt", value="(deformed, distorted, disfigured:1.3), poorly drawn, bad anatomy, wrong anatomy, extra limb, missing limb, floating limbs, (mutated hands and fingers:1.4), disconnected limbs, mutation, mutated, ugly, disgusting, blurry, amputation", visible=True, ) seed = gr.Slider( label="Seed", minimum=0, maximum=MAX_SEED, step=1, value=0, ) randomize_seed = gr.Checkbox(label="Randomize seed", value=True) with gr.Row(visible=True): width = gr.Slider( label="Width", minimum=512, maximum=MAX_IMAGE_SIZE, step=64, value=1024, ) height = gr.Slider( label="Height", minimum=512, maximum=MAX_IMAGE_SIZE, step=64, value=1024, ) with gr.Row(): guidance_scale = gr.Slider( label="Guidance Scale", minimum=0.1, maximum=6, step=0.1, value=3.0, ) num_inference_steps = gr.Slider( label="Number of inference steps", minimum=1, maximum=25, step=1, value=23, ) with gr.Column(scale=1): gr.Examples( examples=examples, inputs=prompt, cache_examples=False, ) use_negative_prompt.change( fn=lambda x: gr.update(visible=x), inputs=use_negative_prompt, outputs=negative_prompt, api_name=False, ) gr.on( triggers=[ prompt.submit, negative_prompt.submit, run_button.click, ], fn=generate, inputs=[ prompt, negative_prompt, use_negative_prompt, seed, width, height, guidance_scale, num_inference_steps, randomize_seed, num_images ], outputs=[result, seed], api_name="run", ) if __name__ == "__main__": demo.queue(max_size=40).launch(ssr_mode=False)