|
|
|
|
|
from __future__ import annotations |
|
|
|
import os |
|
import random |
|
import gc |
|
import toml |
|
import gradio as gr |
|
import numpy as np |
|
import utils |
|
import torch |
|
import json |
|
import PIL.Image |
|
import base64 |
|
import safetensors |
|
from io import BytesIO |
|
from typing import Tuple |
|
import gradio_user_history as gr_user_history |
|
from huggingface_hub import hf_hub_download |
|
from safetensors.torch import load_file |
|
from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer |
|
from lora_diffusers import LoRANetwork, create_network_from_weights |
|
from diffusers.models import AutoencoderKL |
|
from diffusers import ( |
|
LCMScheduler, |
|
StableDiffusionXLPipeline, |
|
StableDiffusionXLImg2ImgPipeline, |
|
DPMSolverMultistepScheduler, |
|
DPMSolverSinglestepScheduler, |
|
KDPM2DiscreteScheduler, |
|
EulerDiscreteScheduler, |
|
EulerAncestralDiscreteScheduler, |
|
HeunDiscreteScheduler, |
|
LMSDiscreteScheduler, |
|
DDIMScheduler, |
|
DEISMultistepScheduler, |
|
UniPCMultistepScheduler, |
|
) |
|
|
|
DESCRIPTION = "Animagine XL 2.0" |
|
|
|
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" |
|
ENABLE_REFINER_PROMPT = os.getenv("ENABLE_REFINER_PROMPT") == "1" |
|
MAX_SEED = np.iinfo(np.int32).max |
|
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" |
|
|
|
MODEL = os.getenv("MODEL", "Linaqruf/animagine-xl-2.0") |
|
|
|
torch.backends.cudnn.deterministic = True |
|
torch.backends.cudnn.benchmark = False |
|
|
|
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu") |
|
|
|
if torch.cuda.is_available(): |
|
if ENABLE_REFINER_PROMPT: |
|
tokenizer = AutoTokenizer.from_pretrained("isek-ai/SDPrompt-RetNet-300M") |
|
tuner = AutoModelForCausalLM.from_pretrained( |
|
"isek-ai/SDPrompt-RetNet-300M", |
|
trust_remote_code=True, |
|
).to(device) |
|
vae = AutoencoderKL.from_pretrained( |
|
"madebyollin/sdxl-vae-fp16-fix", |
|
torch_dtype=torch.float16, |
|
) |
|
pipe = StableDiffusionXLPipeline.from_pretrained( |
|
MODEL, |
|
vae=vae, |
|
torch_dtype=torch.float16, |
|
custom_pipeline="lpw_stable_diffusion_xl", |
|
use_safetensors=True, |
|
use_auth_token=HF_TOKEN, |
|
variant="fp16", |
|
) |
|
|
|
if ENABLE_CPU_OFFLOAD: |
|
pipe.enable_model_cpu_offload() |
|
else: |
|
pipe.to(device) |
|
if USE_TORCH_COMPILE: |
|
pipe.unet = torch.compile(pipe.unet, mode="reduce-overhead", fullgraph=True) |
|
else: |
|
pipe = None |
|
|
|
|
|
def randomize_seed_fn(seed: int, randomize_seed: bool) -> int: |
|
if randomize_seed: |
|
seed = random.randint(0, MAX_SEED) |
|
return seed |
|
|
|
|
|
def seed_everything(seed): |
|
torch.manual_seed(seed) |
|
torch.cuda.manual_seed_all(seed) |
|
np.random.seed(seed) |
|
random.seed(seed) |
|
generator = torch.Generator() |
|
generator.manual_seed(seed) |
|
return generator |
|
|
|
|
|
def get_image_path(base_path: str): |
|
extensions = [".jpg", ".jpeg", ".png", ".bmp", ".gif"] |
|
for ext in extensions: |
|
image_path = base_path + ext |
|
if os.path.exists(image_path): |
|
return image_path |
|
return None |
|
|
|
|
|
def update_lcm_parameter(enable_lcm: bool = False): |
|
if enable_lcm: |
|
return (2, 8, gr.update(value="LCM"), gr.update(choices=["LCM"])) |
|
else: |
|
return (12, 50, gr.update(value="Euler a"), gr.update(choices=sampler_list)) |
|
|
|
|
|
def update_selection(selected_state: gr.SelectData): |
|
lora_repo = sdxl_loras[selected_state.index]["repo"] |
|
lora_weight = sdxl_loras[selected_state.index]["multiplier"] |
|
updated_selected_info = f"{lora_repo}" |
|
|
|
return ( |
|
updated_selected_info, |
|
selected_state, |
|
lora_weight, |
|
) |
|
|
|
|
|
def parse_aspect_ratio(aspect_ratio): |
|
if aspect_ratio == "Custom": |
|
return None, None |
|
width, height = aspect_ratio.split(" x ") |
|
return int(width), int(height) |
|
|
|
|
|
def aspect_ratio_handler(aspect_ratio, custom_width, custom_height): |
|
if aspect_ratio == "Custom": |
|
return custom_width, custom_height |
|
else: |
|
width, height = parse_aspect_ratio(aspect_ratio) |
|
return width, height |
|
|
|
|
|
def create_network(text_encoders, unet, state_dict, multiplier, device): |
|
network = create_network_from_weights( |
|
text_encoders, |
|
unet, |
|
state_dict, |
|
multiplier, |
|
) |
|
network.load_state_dict(state_dict) |
|
network.to(device, dtype=unet.dtype) |
|
network.apply_to(multiplier=multiplier) |
|
|
|
return network |
|
|
|
|
|
def get_scheduler(scheduler_config, name): |
|
scheduler_map = { |
|
"DPM++ 2M Karras": lambda: DPMSolverMultistepScheduler.from_config( |
|
scheduler_config, use_karras_sigmas=True |
|
), |
|
"DPM++ SDE Karras": lambda: DPMSolverSinglestepScheduler.from_config( |
|
scheduler_config, use_karras_sigmas=True |
|
), |
|
"DPM++ 2M SDE Karras": lambda: DPMSolverMultistepScheduler.from_config( |
|
scheduler_config, use_karras_sigmas=True, algorithm_type="sde-dpmsolver++" |
|
), |
|
"Euler": lambda: EulerDiscreteScheduler.from_config(scheduler_config), |
|
"Euler a": lambda: EulerAncestralDiscreteScheduler.from_config( |
|
scheduler_config |
|
), |
|
"DDIM": lambda: DDIMScheduler.from_config(scheduler_config), |
|
"LCM": lambda: LCMScheduler.from_config(scheduler_config), |
|
} |
|
return scheduler_map.get(name, lambda: None)() |
|
|
|
|
|
def free_memory(): |
|
torch.cuda.empty_cache() |
|
gc.collect() |
|
|
|
|
|
def preprocess_prompt( |
|
style_dict, |
|
style_name: str, |
|
positive: str, |
|
negative: str = "", |
|
) -> Tuple[str, str]: |
|
p, n = style_dict.get(style_name, styles["(None)"]) |
|
|
|
return p.format(prompt=positive), n + negative |
|
|
|
|
|
def common_upscale(samples, width, height, upscale_method): |
|
return torch.nn.functional.interpolate( |
|
samples, size=(height, width), mode=upscale_method |
|
) |
|
|
|
|
|
def upscale(samples, upscale_method, scale_by): |
|
width = round(samples.shape[3] * scale_by) |
|
height = round(samples.shape[2] * scale_by) |
|
s = common_upscale(samples, width, height, upscale_method) |
|
return s |
|
|
|
|
|
def prompt_completion( |
|
input_text, |
|
max_new_tokens=128, |
|
do_sample=True, |
|
temperature=1.0, |
|
top_p=0.95, |
|
top_k=20, |
|
repetition_penalty=1.2, |
|
num_beams=1, |
|
): |
|
try: |
|
if input_text.strip() == "": |
|
return "" |
|
|
|
inputs = tokenizer( |
|
f"<s>{input_text}", return_tensors="pt", add_special_tokens=False |
|
)["input_ids"].to(device) |
|
|
|
result = tuner.generate( |
|
inputs, |
|
max_new_tokens=max_new_tokens, |
|
do_sample=do_sample, |
|
temperature=temperature, |
|
top_p=top_p, |
|
top_k=top_k, |
|
repetition_penalty=repetition_penalty, |
|
num_beams=num_beams, |
|
) |
|
|
|
return tokenizer.batch_decode(result, skip_special_tokens=True)[0] |
|
|
|
except Exception as e: |
|
print(f"An error occured: {e}") |
|
raise |
|
|
|
finally: |
|
free_memory() |
|
|
|
|
|
def load_and_convert_thumbnail(model_path: str): |
|
with safetensors.safe_open(model_path, framework="pt") as f: |
|
metadata = f.metadata() |
|
if "modelspec.thumbnail" in metadata: |
|
base64_data = metadata["modelspec.thumbnail"] |
|
prefix, encoded = base64_data.split(",", 1) |
|
image_data = base64.b64decode(encoded) |
|
image = PIL.Image.open(BytesIO(image_data)) |
|
return image |
|
return None |
|
|
|
|
|
def generate( |
|
prompt: str, |
|
negative_prompt: str = "", |
|
seed: int = 0, |
|
custom_width: int = 1024, |
|
custom_height: int = 1024, |
|
guidance_scale: float = 12.0, |
|
num_inference_steps: int = 50, |
|
use_lora: bool = False, |
|
lora_weight: float = 1.0, |
|
selected_state: str = "", |
|
enable_lcm: bool = False, |
|
sampler: str = "Euler a", |
|
aspect_ratio_selector: str = "1024 x 1024", |
|
style_selector: str = "(None)", |
|
quality_selector: str = "Standard", |
|
use_upscaler: bool = False, |
|
upscaler_strength: float = 0.5, |
|
upscale_by: float = 1.5, |
|
refine_prompt: bool = False, |
|
profile: gr.OAuthProfile | None = None, |
|
progress=gr.Progress(track_tqdm=True), |
|
) -> PIL.Image.Image: |
|
generator = seed_everything(seed) |
|
|
|
network = None |
|
network_state = {"current_lora": None, "multiplier": None} |
|
adapter_id = "Linaqruf/lcm-lora-sdxl-rank1" |
|
|
|
width, height = aspect_ratio_handler( |
|
aspect_ratio_selector, |
|
custom_width, |
|
custom_height, |
|
) |
|
|
|
if ENABLE_REFINER_PROMPT: |
|
if refine_prompt: |
|
if not prompt: |
|
prompt = random.choice(["1girl, solo", "1boy, solo"]) |
|
prompt = prompt_completion(prompt) |
|
|
|
prompt, negative_prompt = preprocess_prompt( |
|
quality_prompt, quality_selector, prompt, negative_prompt |
|
) |
|
prompt, negative_prompt = preprocess_prompt( |
|
styles, style_selector, prompt, negative_prompt |
|
) |
|
|
|
if width % 8 != 0: |
|
width = width - (width % 8) |
|
if height % 8 != 0: |
|
height = height - (height % 8) |
|
if use_lora: |
|
if not selected_state: |
|
raise Exception("You must Select a LoRA") |
|
repo_name = sdxl_loras[selected_state.index]["repo"] |
|
full_path_lora = saved_names[selected_state.index] |
|
weight_name = sdxl_loras[selected_state.index]["weights"] |
|
|
|
lora_sd = load_file(full_path_lora) |
|
text_encoders = [pipe.text_encoder, pipe.text_encoder_2] |
|
|
|
if network_state["current_lora"] != repo_name: |
|
network = create_network( |
|
text_encoders, |
|
pipe.unet, |
|
lora_sd, |
|
lora_weight, |
|
device, |
|
) |
|
network_state["current_lora"] = repo_name |
|
network_state["multiplier"] = lora_weight |
|
elif network_state["multiplier"] != lora_weight: |
|
network = create_network( |
|
text_encoders, |
|
pipe.unet, |
|
lora_sd, |
|
lora_weight, |
|
device, |
|
) |
|
network_state["multiplier"] = lora_weight |
|
else: |
|
if network: |
|
network.unapply_to() |
|
network = None |
|
network_state = { |
|
"current_lora": None, |
|
"multiplier": None, |
|
} |
|
|
|
if enable_lcm: |
|
pipe.load_lora_weights(adapter_id) |
|
|
|
backup_scheduler = pipe.scheduler |
|
pipe.scheduler = 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, |
|
"enable_lcm": enable_lcm, |
|
"sdxl_style": style_selector, |
|
"quality_tags": quality_selector, |
|
"refine_prompt": refine_prompt, |
|
} |
|
|
|
if use_lora: |
|
metadata["use_lora"] = {"selected_lora": repo_name, "multiplier": lora_weight} |
|
else: |
|
metadata["use_lora"] = None |
|
|
|
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 |
|
|
|
print(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 = upscale(latents, "nearest-exact", upscale_by) |
|
image = 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[0] |
|
else: |
|
image = 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[0] |
|
if network: |
|
network.unapply_to() |
|
network = None |
|
if profile is not None: |
|
gr_user_history.save_image( |
|
label=prompt, |
|
image=image, |
|
profile=profile, |
|
metadata=metadata, |
|
) |
|
return image, metadata |
|
except Exception as e: |
|
print(f"An error occured: {e}") |
|
raise |
|
finally: |
|
if network: |
|
network.unapply_to() |
|
network = None |
|
if use_lora: |
|
del lora_sd, text_encoders |
|
if enable_lcm: |
|
pipe.unload_lora_weights() |
|
if use_upscaler: |
|
del upscaler_pipe |
|
pipe.scheduler = backup_scheduler |
|
free_memory() |
|
|
|
|
|
examples = [ |
|
"face focus, cute, 1girl, green hair, sweater, looking at viewer, upper body, beanie, outdoors, night, turtleneck", |
|
"face focus, bishounen, 1boy, green hair, sweater, looking at viewer, upper body, beanie, outdoors, night, turtleneck", |
|
"face focus, fu xuan, 1girl, solo, yellow eyes, dress, looking at viewer, hair rings, bare shoulders, long hair, hair ornament, purple hair, bangs, forehead jewel, frills, tassel, jewelry, pink hair", |
|
"face focus, bishounen, 1boy, zhongli, looking at viewer, upper body, outdoors, night", |
|
"a girl with mesmerizing blue eyes peers at the viewer. Her long, white hair flows gracefully, adorned with stunning blue butterfly hair ornaments", |
|
] |
|
|
|
quality_prompt_list = [ |
|
{ |
|
"name": "(None)", |
|
"prompt": "{prompt}", |
|
"negative_prompt": "", |
|
}, |
|
{ |
|
"name": "Standard", |
|
"prompt": "masterpiece, best quality, {prompt}", |
|
"negative_prompt": "lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry", |
|
}, |
|
{ |
|
"name": "Light", |
|
"prompt": "(masterpiece), best quality, expressive eyes, perfect face, {prompt}", |
|
"negative_prompt": "(low quality, worst quality:1.2), 3d, watermark, signature, ugly, poorly drawn", |
|
}, |
|
{ |
|
"name": "Heavy", |
|
"prompt": "(masterpiece), (best quality), (ultra-detailed), {prompt}, illustration, disheveled hair, detailed eyes, perfect composition, moist skin, intricate details, earrings", |
|
"negative_prompt": "longbody, lowres, bad anatomy, bad hands, missing fingers, pubic hair, extra digit, fewer digits, cropped, worst quality, low quality", |
|
}, |
|
] |
|
|
|
sampler_list = [ |
|
"DPM++ 2M Karras", |
|
"DPM++ SDE Karras", |
|
"DPM++ 2M SDE Karras", |
|
"Euler", |
|
"Euler a", |
|
"DDIM", |
|
] |
|
|
|
aspect_ratios = [ |
|
"1024 x 1024", |
|
"1152 x 896", |
|
"896 x 1152", |
|
"1216 x 832", |
|
"832 x 1216", |
|
"1344 x 768", |
|
"768 x 1344", |
|
"1536 x 640", |
|
"640 x 1536", |
|
"Custom", |
|
] |
|
|
|
style_list = [ |
|
{ |
|
"name": "(None)", |
|
"prompt": "{prompt}", |
|
"negative_prompt": "", |
|
}, |
|
{ |
|
"name": "Cinematic", |
|
"prompt": "cinematic still {prompt} . emotional, harmonious, vignette, highly detailed, high budget, bokeh, cinemascope, moody, epic, gorgeous, film grain, grainy", |
|
"negative_prompt": "cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured", |
|
}, |
|
{ |
|
"name": "Photographic", |
|
"prompt": "cinematic photo {prompt} . 35mm photograph, film, bokeh, professional, 4k, highly detailed", |
|
"negative_prompt": "drawing, painting, crayon, sketch, graphite, impressionist, noisy, blurry, soft, deformed, ugly", |
|
}, |
|
{ |
|
"name": "Anime", |
|
"prompt": "anime artwork {prompt} . anime style, key visual, vibrant, studio anime, highly detailed", |
|
"negative_prompt": "photo, deformed, black and white, realism, disfigured, low contrast", |
|
}, |
|
{ |
|
"name": "Manga", |
|
"prompt": "manga style {prompt} . vibrant, high-energy, detailed, iconic, Japanese comic style", |
|
"negative_prompt": "ugly, deformed, noisy, blurry, low contrast, realism, photorealistic, Western comic style", |
|
}, |
|
{ |
|
"name": "Digital Art", |
|
"prompt": "concept art {prompt} . digital artwork, illustrative, painterly, matte painting, highly detailed", |
|
"negative_prompt": "photo, photorealistic, realism, ugly", |
|
}, |
|
{ |
|
"name": "Pixel art", |
|
"prompt": "pixel-art {prompt} . low-res, blocky, pixel art style, 8-bit graphics", |
|
"negative_prompt": "sloppy, messy, blurry, noisy, highly detailed, ultra textured, photo, realistic", |
|
}, |
|
{ |
|
"name": "Fantasy art", |
|
"prompt": "ethereal fantasy concept art of {prompt} . magnificent, celestial, ethereal, painterly, epic, majestic, magical, fantasy art, cover art, dreamy", |
|
"negative_prompt": "photographic, realistic, realism, 35mm film, dslr, cropped, frame, text, deformed, glitch, noise, noisy, off-center, deformed, cross-eyed, closed eyes, bad anatomy, ugly, disfigured, sloppy, duplicate, mutated, black and white", |
|
}, |
|
{ |
|
"name": "Neonpunk", |
|
"prompt": "neonpunk style {prompt} . cyberpunk, vaporwave, neon, vibes, vibrant, stunningly beautiful, crisp, detailed, sleek, ultramodern, magenta highlights, dark purple shadows, high contrast, cinematic, ultra detailed, intricate, professional", |
|
"negative_prompt": "painting, drawing, illustration, glitch, deformed, mutated, cross-eyed, ugly, disfigured", |
|
}, |
|
{ |
|
"name": "3D Model", |
|
"prompt": "professional 3d model {prompt} . octane render, highly detailed, volumetric, dramatic lighting", |
|
"negative_prompt": "ugly, deformed, noisy, low poly, blurry, painting", |
|
}, |
|
] |
|
|
|
thumbnail_cache = {} |
|
|
|
with open("lora.toml", "r") as file: |
|
data = toml.load(file) |
|
|
|
sdxl_loras = [] |
|
saved_names = [] |
|
for item in data["data"]: |
|
model_path = hf_hub_download(item["repo"], item["weights"], token=HF_TOKEN) |
|
saved_names.append(model_path) |
|
|
|
if model_path not in thumbnail_cache: |
|
thumbnail_image = load_and_convert_thumbnail(model_path) |
|
thumbnail_cache[model_path] = thumbnail_image |
|
else: |
|
thumbnail_image = thumbnail_cache[model_path] |
|
|
|
sdxl_loras.append( |
|
{ |
|
"image": thumbnail_image, |
|
"title": item["title"], |
|
"repo": item["repo"], |
|
"weights": item["weights"], |
|
"multiplier": item.get("multiplier", "1.0"), |
|
} |
|
) |
|
|
|
styles = {k["name"]: (k["prompt"], k["negative_prompt"]) for k in style_list} |
|
quality_prompt = { |
|
k["name"]: (k["prompt"], k["negative_prompt"]) for k in quality_prompt_list |
|
} |
|
|
|
|
|
|
|
|
|
|
|
|
|
with gr.Blocks(css="style.css", theme="NoCrypt/[email protected]") as demo: |
|
title = gr.HTML( |
|
f"""<h1><span>{DESCRIPTION}</span></h1>""", |
|
elem_id="title", |
|
) |
|
gr.Markdown( |
|
f"""Gradio demo for [Linaqruf/animagine-xl-2.0](https://huggingface.co/Linaqruf/animagine-xl-2.0)""", |
|
elem_id="subtitle", |
|
) |
|
gr.DuplicateButton( |
|
value="Duplicate Space for private use", |
|
elem_id="duplicate-button", |
|
visible=os.getenv("SHOW_DUPLICATE_BUTTON") == "1", |
|
) |
|
selected_state = gr.State() |
|
with gr.Row(): |
|
with gr.Column(scale=2): |
|
with gr.Tab("Txt2img"): |
|
with gr.Group(): |
|
prompt = gr.Text( |
|
label="Prompt", |
|
max_lines=5, |
|
placeholder="Enter your prompt", |
|
) |
|
negative_prompt = gr.Text( |
|
label="Negative Prompt", |
|
max_lines=5, |
|
placeholder="Enter a negative prompt", |
|
) |
|
with gr.Accordion(label="Quality Prompt Presets", open=False): |
|
quality_selector = gr.Dropdown( |
|
label="Quality Prompt Presets", |
|
show_label=False, |
|
interactive=True, |
|
choices=list(quality_prompt.keys()), |
|
value="Standard", |
|
) |
|
with gr.Row(): |
|
enable_lcm = gr.Checkbox(label="Enable LCM", value=False) |
|
use_lora = gr.Checkbox(label="Use LoRA", value=False) |
|
refine_prompt = gr.Checkbox( |
|
label="Refine prompt", |
|
value=False, |
|
visible=ENABLE_REFINER_PROMPT, |
|
) |
|
with gr.Group(visible=False) as lora_group: |
|
selector_info = gr.Text( |
|
label="Selected LoRA", |
|
max_lines=1, |
|
value="No LoRA selected.", |
|
) |
|
lora_selection = gr.Gallery( |
|
value=[(item["image"], item["title"]) for item in sdxl_loras], |
|
label="Animagine XL 2.0 LoRA", |
|
show_label=False, |
|
columns=2, |
|
show_share_button=False, |
|
) |
|
lora_weight = gr.Slider( |
|
label="Multiplier", |
|
minimum=-2, |
|
maximum=2, |
|
step=0.05, |
|
value=1, |
|
) |
|
with gr.Tab("Advanced Settings"): |
|
with gr.Group(): |
|
style_selector = gr.Radio( |
|
label="Style Preset", |
|
container=True, |
|
interactive=True, |
|
choices=list(styles.keys()), |
|
value="(None)", |
|
) |
|
with gr.Group(): |
|
aspect_ratio_selector = gr.Radio( |
|
label="Aspect Ratio", |
|
choices=aspect_ratios, |
|
value="1024 x 1024", |
|
container=True, |
|
) |
|
with gr.Group(): |
|
use_upscaler = gr.Checkbox(label="Use Upscaler", value=False) |
|
with gr.Row() as upscaler_row: |
|
upscaler_strength = gr.Slider( |
|
label="Strength", |
|
minimum=0, |
|
maximum=1, |
|
step=0.05, |
|
value=0.55, |
|
visible=False, |
|
) |
|
upscale_by = gr.Slider( |
|
label="Upscale by", |
|
minimum=1, |
|
maximum=1.5, |
|
step=0.1, |
|
value=1.5, |
|
visible=False, |
|
) |
|
with gr.Group(visible=False) as custom_resolution: |
|
with gr.Row(): |
|
custom_width = gr.Slider( |
|
label="Width", |
|
minimum=MIN_IMAGE_SIZE, |
|
maximum=MAX_IMAGE_SIZE, |
|
step=8, |
|
value=1024, |
|
) |
|
custom_height = gr.Slider( |
|
label="Height", |
|
minimum=MIN_IMAGE_SIZE, |
|
maximum=MAX_IMAGE_SIZE, |
|
step=8, |
|
value=1024, |
|
) |
|
with gr.Group(): |
|
sampler = gr.Dropdown( |
|
label="Sampler", |
|
choices=sampler_list, |
|
interactive=True, |
|
value="Euler a", |
|
) |
|
with gr.Group(): |
|
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.Group(): |
|
with gr.Row(): |
|
guidance_scale = gr.Slider( |
|
label="Guidance scale", |
|
minimum=1, |
|
maximum=20, |
|
step=0.1, |
|
value=12.0, |
|
) |
|
num_inference_steps = gr.Slider( |
|
label="Number of inference steps", |
|
minimum=1, |
|
maximum=100, |
|
step=1, |
|
value=50, |
|
) |
|
|
|
with gr.Tab("Past Generation"): |
|
gr_user_history.render() |
|
with gr.Column(scale=3): |
|
with gr.Blocks(): |
|
run_button = gr.Button("Generate", variant="primary") |
|
result = gr.Image(label="Result", show_label=False) |
|
with gr.Accordion(label="Generation Parameters", open=False): |
|
gr_metadata = gr.JSON(label="Metadata", show_label=False) |
|
gr.Examples( |
|
examples=examples, |
|
inputs=prompt, |
|
outputs=[result, gr_metadata], |
|
fn=generate, |
|
cache_examples=CACHE_EXAMPLES, |
|
) |
|
|
|
lora_selection.select( |
|
update_selection, |
|
outputs=[ |
|
selector_info, |
|
selected_state, |
|
lora_weight, |
|
], |
|
queue=False, |
|
show_progress=False, |
|
) |
|
enable_lcm.change( |
|
update_lcm_parameter, |
|
inputs=enable_lcm, |
|
outputs=[ |
|
guidance_scale, |
|
num_inference_steps, |
|
sampler, |
|
sampler, |
|
], |
|
queue=False, |
|
api_name=False, |
|
) |
|
use_lora.change( |
|
fn=lambda x: gr.update(visible=x), |
|
inputs=use_lora, |
|
outputs=lora_group, |
|
queue=False, |
|
api_name=False, |
|
) |
|
use_upscaler.change( |
|
fn=lambda x: [gr.update(visible=x), gr.update(visible=x)], |
|
inputs=use_upscaler, |
|
outputs=[upscaler_strength, upscale_by], |
|
queue=False, |
|
api_name=False, |
|
) |
|
aspect_ratio_selector.change( |
|
fn=lambda x: gr.update(visible=x == "Custom"), |
|
inputs=aspect_ratio_selector, |
|
outputs=custom_resolution, |
|
queue=False, |
|
api_name=False, |
|
) |
|
|
|
inputs = [ |
|
prompt, |
|
negative_prompt, |
|
seed, |
|
custom_width, |
|
custom_height, |
|
guidance_scale, |
|
num_inference_steps, |
|
use_lora, |
|
lora_weight, |
|
selected_state, |
|
enable_lcm, |
|
sampler, |
|
aspect_ratio_selector, |
|
style_selector, |
|
quality_selector, |
|
use_upscaler, |
|
upscaler_strength, |
|
upscale_by, |
|
refine_prompt, |
|
] |
|
|
|
prompt.submit( |
|
fn=randomize_seed_fn, |
|
inputs=[seed, randomize_seed], |
|
outputs=seed, |
|
queue=False, |
|
api_name=False, |
|
).then( |
|
fn=generate, |
|
inputs=inputs, |
|
outputs=result, |
|
api_name="run", |
|
) |
|
negative_prompt.submit( |
|
fn=randomize_seed_fn, |
|
inputs=[seed, randomize_seed], |
|
outputs=seed, |
|
queue=False, |
|
api_name=False, |
|
).then( |
|
fn=generate, |
|
inputs=inputs, |
|
outputs=result, |
|
api_name=False, |
|
) |
|
run_button.click( |
|
fn=randomize_seed_fn, |
|
inputs=[seed, randomize_seed], |
|
outputs=seed, |
|
queue=False, |
|
api_name=False, |
|
).then( |
|
fn=generate, |
|
inputs=inputs, |
|
outputs=[result, gr_metadata], |
|
api_name=False, |
|
) |
|
demo.queue(max_size=20).launch(debug=IS_COLAB, share=IS_COLAB) |
|
|
