gradio-app.py

import time
import gradio as gr
from PIL import Image
import torch
from diffusers import DiffusionPipeline, AutoencoderTiny
import os, sys

import subprocess


subprocess.check_call([sys.executable, "-m", "pip", "install", 'stable-fast'])

SAFETY_CHECKER = os.environ.get("SAFETY_CHECKER", None)
TORCH_COMPILE = os.environ.get("TORCH_COMPILE", None)
# I have no idea where the env is config'ed.  Thus:
TORCH_COMPILE = True

if SAFETY_CHECKER:
    pipe = DiffusionPipeline.from_pretrained(
        "SimianLuo/LCM_Dreamshaper_v7",
        custom_pipeline="lcm_txt2img",
        scheduler=None,
    )
else:
    pipe = DiffusionPipeline.from_pretrained(
        "SimianLuo/LCM_Dreamshaper_v7",
        custom_pipeline="lcm_txt2img",
        scheduler=None,
        safety_checker=None,
    )
pipe.to(device="cuda", dtype=torch.float16)
pipe.vae = AutoencoderTiny.from_pretrained(
    "madebyollin/taesd", device="cuda", torch_dtype=torch.float16
)
pipe.vae = pipe.vae.cuda()
pipe.unet.to(memory_format=torch.channels_last)
pipe.set_progress_bar_config(disable=True)

if TORCH_COMPILE:
    #optmode = 'max-autotune'
    #optmode = 'reduce-overhead'
    #pipe.text_encoder = torch.compile(pipe.text_encoder, mode=optmode)
    #pipe.tokenizer = torch.compile(pipe.tokenizer, mode=optmode)
    #pipe.unet = torch.compile(pipe.unet, mode=optmode)
    #pipe.vae = torch.compile(pipe.vae, mode=optmode)
    doCompile = True
    if doCompile:
        config = CompilationConfig.Default()

        try:
            import xformers
            config.enable_xformers = True
        except ImportError:
            print('xformers not installed, skipping')
        try:
            import triton
            config.enable_triton = True
        except ImportError:
            print('Triton not installed, skipping')

        config.enable_cuda_graph = True
        config.enable_jit = True
        config.enable_jit_freeze = True
        config.trace_scheduler = False#True   CHECK THIS AGAIN
        config.enable_cnn_optimization = True
        config.preserve_parameters = False
        config.prefer_lowp_gemm = True
        config.enable_fused_linear_geglu = True

        torch.jit.optimize_for_inference = True
        torch.jit.enable_onednn_fusion = True
        torch.jit.set_fusion_strategy([('STATIC', 1), ('DYNAMIC', 1)])

        for p in pipe.text_encoder.parameters(): p.requires_grad=False
        for p in pipe.vae.decoder.parameters(): p.requires_grad=False
        for p in pipe.unet.parameters(): p.requires_grad=False

def predict(prompt1, prompt2, merge_ratio, guidance, steps, sharpness, seed=1231231):
    torch.manual_seed(seed)
    tm0 = time.time()
    results = pipe(
        prompt1=prompt1,
        prompt2=prompt2,
        sv=merge_ratio,
        sharpness=sharpness,
        width=640,
        height=640,
        num_inference_steps=steps,
        guidance_scale=guidance,
        lcm_origin_steps=50,
        output_type="pil",
        # return_dict=False,
    )
    torch.cuda.synchronize()
    tmval = f"time = {time.time()-tm0}"
    print(f"time = {time.time()-tm0}")

    nsfw_content_detected = (
        results.nsfw_content_detected[0]
        if "nsfw_content_detected" in results
        else False
    )
    if nsfw_content_detected:
        raise gr.Error("NSFW content detected. Please try another prompt.")
    return results.images[0], tmval


css = """
#container{
    margin: 0 auto;
    max-width: 80rem;
}
#intro{
    max-width: 32rem;
    text-align: center;
    margin: 0 auto;
}
"""
with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="container"):
        gr.Markdown(
            """# SDZoom

    Welcome to sdzoom, a testbed application designed for optimizing and experimenting with various 
    configurations to achieve the fastest Stable Diffusion (SD) pipelines.
    RTSD leverages the expertise provided by Latent Consistency Models (LCM). For more information about LCM,
    visit their website at [Latent Consistency Models](https://latent-consistency-models.github.io/).

    """,
            elem_id="intro",
        )
        with gr.Row():
            with gr.Column():
                image = gr.Image(type="pil")
            with gr.Column():
                merge_ratio = gr.Slider(
                    value=50, minimum=1, maximum=100, step=1, label="Merge Ratio"
                )
                guidance = gr.Slider(
                    label="Guidance", minimum=1, maximum=50, value=10.0, step=0.01
                )
                steps = gr.Slider(label="Steps", value=4, minimum=2, maximum=20, step=1)
                sharpness = gr.Slider(
                    value=1.0, minimum=0, maximum=1, step=0.001, label="Sharpness"
                )
                seed = gr.Slider(
                    randomize=True, minimum=0, maximum=12013012031030, label="Seed"
                )
                prompt1 = gr.Textbox(label="Prompt 1")
                prompt2 = gr.Textbox(label="Prompt 2")
                msg = gr.Textbox(label="Message", interactive=False)
                generate_bt = gr.Button("Generate")

        inputs = [prompt1, prompt2, merge_ratio, guidance, steps, sharpness, seed]
        gr.Examples(
            examples=[
                ["Elon Musk", "Mark Zuckerberg", 50, 10.0, 4, 1.0, 1231231],
                ["Elon Musk", "Bill Gates", 50, 10.0, 4, 1.0, 53453],
                [
                    "Asian women, intricate jewlery in her hair, 8k",
                    "Tom Cruise, intricate jewlery in her hair, 8k",
                    50,
                    10.0,
                    4,
                    1.0,
                    542343,
                ],
            ],
            fn=predict,
            inputs=inputs,
            outputs=[image, msg],
        )
        generate_bt.click(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        seed.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        merge_ratio.change(
            fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False
        )
        guidance.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        steps.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        sharpness.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        prompt1.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)
        prompt2.change(fn=predict, inputs=inputs, outputs=[image, msg], show_progress=False)

demo.queue()
if __name__ == "__main__":
    demo.launch()