#!/usr/bin/env python
from __future__ import annotations
import os, random, glob, re
from datetime import datetime
import gradio as gr
import numpy as np
import PIL.Image
import spaces
import torch
from diffusers import AutoencoderKL, DiffusionPipeline
DESCRIPTION = """
# OpenDalle 1.1 with Gallery
**Demo by [mrfakename](https://mrfake.name/) - [Twitter](https://twitter.com/realmrfakename) - [GitHub](https://github.com/fakerybakery/) - [Hugging Face](https://huggingface.co/mrfakename)**
This is a demo of OpenDalle V1.1 by @dataautogpt3.
It's a merge of several different models and is supposed to provide excellent performance. Try it out!
[Not Working?](https://huggingface.co/spaces/mrfakename/OpenDalleV1.1-GPU-Demo/discussions/4)
**The code for this demo is based on [@hysts's SD-XL demo](https://huggingface.co/spaces/hysts/SD-XL) running on a A10G GPU.**
**NOTE: The model is licensed under a non-commercial license**
Also see [OpenDalle Original Demo](https://huggingface.co/spaces/mrfakename/OpenDalle-GPU-Demo/)
"""
if not torch.cuda.is_available():
DESCRIPTION += "\nRunning on CPU 🥶 This demo does not work on CPU. Please use the online demo instead
"
MAX_SEED = np.iinfo(np.int32).max
CACHE_EXAMPLES = torch.cuda.is_available() and os.getenv("CACHE_EXAMPLES", "0") == "1"
MAX_IMAGE_SIZE = int(os.getenv("MAX_IMAGE_SIZE", "1024"))
USE_TORCH_COMPILE = os.getenv("USE_TORCH_COMPILE") == "1"
ENABLE_CPU_OFFLOAD = os.getenv("ENABLE_CPU_OFFLOAD") == "1"
ENABLE_REFINER = os.getenv("ENABLE_REFINER", "0") == "1"
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
if torch.cuda.is_available():
vae = AutoencoderKL.from_pretrained("madebyollin/sdxl-vae-fp16-fix", torch_dtype=torch.float16)
pipe = DiffusionPipeline.from_pretrained("dataautogpt3/OpenDalleV1.1", vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
if ENABLE_REFINER:
refiner = DiffusionPipeline.from_pretrained("stabilityai/stable-diffusion-xl-refiner-1.0", vae=vae, torch_dtype=torch.float16, use_safetensors=True, variant="fp16")
if ENABLE_CPU_OFFLOAD:
pipe.enable_model_cpu_offload()
if ENABLE_REFINER: refiner.enable_model_cpu_offload()
else:
pipe.to(device)
if ENABLE_REFINER: refiner.to(device)
if USE_TORCH_COMPILE:
pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True)
if ENABLE_REFINER: refiner.unet = torch.compile(refiner.unet, mode="reduce-overhead", fullgraph=True)
def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
return random.randint(0, MAX_SEED) if randomize_seed else seed
def save_image(image: PIL.Image.Image, prompt: str) -> str:
timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
clean_prompt = re.sub(r'[^\w\-_\. ]', '_', prompt)[:50]
filename = f"{timestamp}_{clean_prompt}.png"
image.save(filename)
return filename
def get_image_gallery():
image_files = glob.glob("*.png")
return sorted([(file, file) for file in image_files], key=lambda x: os.path.getmtime(x[0]), reverse=True)
@spaces.GPU(enable_queue=True)
def generate(prompt: str, negative_prompt: str = "", prompt_2: str = "", negative_prompt_2: str = "", use_negative_prompt: bool = False, use_prompt_2: bool = False, use_negative_prompt_2: bool = False, seed: int = 0, width: int = 1024, height: int = 1024, guidance_scale_base: float = 5.0, guidance_scale_refiner: float = 5.0, num_inference_steps_base: int = 25, num_inference_steps_refiner: int = 25, apply_refiner: bool = False, progress=gr.Progress(track_tqdm=True)) -> PIL.Image.Image:
print(f"** Generating image for: \"{prompt}\" **")
generator = torch.Generator().manual_seed(seed)
if not use_negative_prompt: negative_prompt = None
if not use_prompt_2: prompt_2 = None
if not use_negative_prompt_2: negative_prompt_2 = None
if not apply_refiner:
image = pipe(prompt=prompt, negative_prompt=negative_prompt, prompt_2=prompt_2, negative_prompt_2=negative_prompt_2, width=width, height=height, guidance_scale=guidance_scale_base, num_inference_steps=num_inference_steps_base, generator=generator, output_type="pil").images[0]
else:
latents = pipe(prompt=prompt, negative_prompt=negative_prompt, prompt_2=prompt_2, negative_prompt_2=negative_prompt_2, width=width, height=height, guidance_scale=guidance_scale_base, num_inference_steps=num_inference_steps_base, generator=generator, output_type="latent").images
image = refiner(prompt=prompt, negative_prompt=negative_prompt, prompt_2=prompt_2, negative_prompt_2=negative_prompt_2, guidance_scale=guidance_scale_refiner, num_inference_steps=num_inference_steps_refiner, image=latents, generator=generator).images[0]
filename = save_image(image, prompt)
return image, get_image_gallery()
examples = [f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} painting of a majestic lighthouse on a rocky coast. Use bold brushstrokes and a vibrant color palette to capture the interplay of light and shadow as the lighthouse beam cuts through a stormy night sky.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} still life featuring a pair of vintage eyeglasses. Focus on the intricate details of the frames and lenses, using a warm color scheme to evoke a sense of nostalgia and wisdom.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} depiction of a rustic wooden stool in a sunlit artist's studio. Emphasize the texture of the wood and the interplay of light and shadow, using a mix of earthy tones and highlights.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} scene viewed through an ornate window frame. Contrast the intricate details of the window with a dreamy, soft-focus landscape beyond, using a palette that transitions from cool interior tones to warm exterior hues.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} close-up study of interlaced fingers. Use a monochromatic color scheme to emphasize the form and texture of the hands, with dramatic lighting to create depth and emotion.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} composition featuring a set of dice in motion. Capture the energy and randomness of the throw, using a dynamic color palette and blurred lines to convey movement.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} interpretation of heaven. Create an ethereal atmosphere with soft, billowing clouds and radiant light, using a palette of celestial blues, golds, and whites.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} portrayal of an ancient, mystical gate. Combine architectural details with elements of fantasy, using a rich, jewel-toned palette to create an air of mystery and magic.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} portrait of a curious cat. Focus on capturing the feline's expressive eyes and sleek form, using a mix of bold and subtle colors to bring out the cat's personality.",
f"{random.choice(['Impressionist', 'Cubist', 'Surrealist', 'Abstract Expressionist', 'Pop Art', 'Minimalist', 'Baroque', 'Art Nouveau', 'Pointillist', 'Fauvism'])} abstract representation of toes in sand. Use textured brushstrokes to convey the feeling of warm sand, with a palette inspired by a sun-drenched beach."]
theme = gr.themes.Base(font=[gr.themes.GoogleFont('Libre Franklin'), gr.themes.GoogleFont('Public Sans'), 'system-ui', 'sans-serif'])
with gr.Blocks(css="footer{display:none !important}", theme=theme) as demo:
gr.Markdown(DESCRIPTION)
gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button", visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1")
with gr.Group():
prompt = gr.Text(label="Prompt", show_label=False, max_lines=1, container=False, placeholder="Enter your prompt")
run_button = gr.Button("Generate")
result = gr.Image(label="Result", show_label=False)
gallery = gr.Gallery(label="Generated Images", show_label=True, columns=4, height="auto")
with gr.Accordion("Advanced options", open=False):
with gr.Row():
use_negative_prompt = gr.Checkbox(label="Use negative prompt", value=False)
use_prompt_2 = gr.Checkbox(label="Use prompt 2", value=False)
use_negative_prompt_2 = gr.Checkbox(label="Use negative prompt 2", value=False)
negative_prompt = gr.Text(label="Negative prompt", max_lines=1, placeholder="Enter a negative prompt", visible=False)
prompt_2 = gr.Text(label="Prompt 2", max_lines=1, placeholder="Enter your prompt", visible=False)
negative_prompt_2 = gr.Text(label="Negative prompt 2", max_lines=1, placeholder="Enter a negative prompt", visible=False)
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():
width = gr.Slider(label="Width", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
height = gr.Slider(label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=1024)
apply_refiner = gr.Checkbox(label="Apply refiner", value=False, visible=ENABLE_REFINER)
with gr.Row():
guidance_scale_base = gr.Slider(label="Guidance scale for base", minimum=1, maximum=20, step=0.1, value=5.0)
num_inference_steps_base = gr.Slider(label="Number of inference steps for base", minimum=10, maximum=100, step=1, value=25)
with gr.Row(visible=False) as refiner_params:
guidance_scale_refiner = gr.Slider(label="Guidance scale for refiner", minimum=1, maximum=20, step=0.1, value=5.0)
num_inference_steps_refiner = gr.Slider(label="Number of inference steps for refiner", minimum=10, maximum=100, step=1, value=25)
gr.Examples(examples=examples, inputs=prompt, outputs=[result, gallery], fn=generate, cache_examples=CACHE_EXAMPLES)
use_negative_prompt.change(fn=lambda x: gr.update(visible=x), inputs=use_negative_prompt, outputs=negative_prompt, queue=False, api_name=False)
use_prompt_2.change(fn=lambda x: gr.update(visible=x), inputs=use_prompt_2, outputs=prompt_2, queue=False, api_name=False)
use_negative_prompt_2.change(fn=lambda x: gr.update(visible=x), inputs=use_negative_prompt_2, outputs=negative_prompt_2, queue=False, api_name=False)
apply_refiner.change(fn=lambda x: gr.update(visible=x), inputs=apply_refiner, outputs=refiner_params, queue=False, api_name=False)
gr.on(triggers=[prompt.submit, negative_prompt.submit, prompt_2.submit, negative_prompt_2.submit, run_button.click],
fn=randomize_seed_fn, inputs=[seed, randomize_seed], outputs=seed, queue=False, api_name=False).then(
fn=generate, inputs=[prompt, negative_prompt, prompt_2, negative_prompt_2, use_negative_prompt, use_prompt_2, use_negative_prompt_2,
seed, width, height, guidance_scale_base, guidance_scale_refiner, num_inference_steps_base, num_inference_steps_refiner, apply_refiner],
outputs=[result, gallery], api_name="run")
demo.load(fn=get_image_gallery, outputs=gallery)
if __name__ == "__main__":
demo.queue(max_size=20, api_open=False).launch(show_api=False)