app.py

import gradio as gr
import numpy as np
import random
import torch
from PIL import Image

from diffusers import (
    DiffusionPipeline,
    StableDiffusionControlNetPipeline,
    ControlNetModel
)
from peft import PeftModel

device = "cuda" if torch.cuda.is_available() else "cpu"

LORA_MODEL = "akaUNik/hw5-homm3-lora-15"
LORA_BASE_MODEL = "runwayml/stable-diffusion-v1-5"

# Model list including LoRA model
MODEL_LIST = [
    "runwayml/stable-diffusion-v1-5",
    "stabilityai/sdxl-turbo",
    "stabilityai/stable-diffusion-2-1",
    LORA_MODEL,  # LoRA model option
]

# ControlNet modes list with aliases
CONTROLNET_MODES = {
    "Canny Edge Detection": "lllyasviel/control_v11p_sd15_canny",
    "Pixel to Pixel": "lllyasviel/control_v11e_sd15_ip2p",
    "Inpainting": "lllyasviel/control_v11p_sd15_inpaint",
    "Multi-Level Line Segments": "lllyasviel/control_v11p_sd15_mlsd",
    "Depth Estimation": "lllyasviel/control_v11f1p_sd15_depth",
    "Surface Normal Estimation": "lllyasviel/control_v11p_sd15_normalbae",
    "Image Segmentation": "lllyasviel/control_v11p_sd15_seg",
    "Line Art Generation": "lllyasviel/control_v11p_sd15_lineart",
    "Anime Line Art": "lllyasviel/control_v11p_sd15_lineart_anime",
    "Human Pose Estimation": "lllyasviel/control_v11p_sd15_openpose",
    "Scribble-Based Generation": "lllyasviel/control_v11p_sd15_scribble",
    "Soft Edge Generation": "lllyasviel/control_v11p_sd15_softedge",
    "Image Shuffling": "lllyasviel/control_v11e_sd15_shuffle",
    "Image Tiling": "lllyasviel/control_v11f1e_sd15_tile",
}

if torch.cuda.is_available():
    torch_dtype = torch.float16
else:
    torch_dtype = torch.float32

# Cache to avoid re-initializing pipelines repeatedly
model_cache = {}

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 512

def infer(
    model_id,
    prompt,
    negative_prompt,
    seed,
    randomize_seed,
    width,
    height,
    guidance_scale,
    num_inference_steps,
    lora_scale,
    controlnet_enable,
    controlnet_mode,
    controlnet_strength,
    controlnet_image,
    ip_adapter_enable,
    ip_adapter_scale,
    ip_adapter_image,
    progress=gr.Progress(track_tqdm=True),
):
    if randomize_seed:
        seed = random.randint(0, MAX_SEED)
    generator = torch.Generator(device=device).manual_seed(seed)

    # Cache
    # if (model_id, controlnet_enable, controlnet_image, controlnet_mode) in model_cache:
      # pipe = model_cache[(model_id, controlnet_enable, controlnet_image, controlnet_mode)]
    # else:

    pipe = None
    if controlnet_enable and controlnet_image:
        controlnet_model = ControlNetModel.from_pretrained(
            CONTROLNET_MODES.get(controlnet_mode),
            torch_dtype=torch_dtype
        )
        if model_id == LORA_MODEL:
            pipe = StableDiffusionControlNetPipeline.from_pretrained(
                LORA_BASE_MODEL,
                controlnet=controlnet_model,
                torch_dtype=torch_dtype
            )
        else:
            pipe = StableDiffusionControlNetPipeline.from_pretrained(
                model_id,
                controlnet=controlnet_model,
                torch_dtype=torch_dtype
            )
    else:
        if model_id == LORA_MODEL:

          # Use the specified base model for your LoRA adapter.
          pipe = DiffusionPipeline.from_pretrained(
              LORA_BASE_MODEL,
              torch_dtype=torch_dtype
          )

          # Load the LoRA weights
          pipe.unet = PeftModel.from_pretrained(
              pipe.unet,
              model_id,
              subfolder="unet",
              torch_dtype=torch_dtype
          )
          pipe.text_encoder = PeftModel.from_pretrained(
              pipe.text_encoder,
              model_id,
              subfolder="text_encoder",
              torch_dtype=torch_dtype
          )
        else:
            pipe = DiffusionPipeline.from_pretrained(
                model_id,
                torch_dtype=torch_dtype
            )

    if ip_adapter_enable:
      pipe.load_ip_adapter(
          "h94/IP-Adapter",
          subfolder="models",
          weight_name="ip-adapter-plus_sd15.bin"
      )
      pipe.set_ip_adapter_scale(ip_adapter_scale)

    pipe.safety_checker = None
    pipe.to(device)
    # model_cache[(model_id, controlnet_enable, controlnet_image, controlnet_mode)] = pipe

    image = pipe(
      prompt=prompt,
      image=controlnet_image if controlnet_enable else None,
      negative_prompt=negative_prompt,
      guidance_scale=guidance_scale,
      num_inference_steps=num_inference_steps,
      width=width,
      height=height,
      generator=generator,
      cross_attention_kwargs={"scale": lora_scale},
      controlnet_conditioning_scale=controlnet_strength,
      ip_adapter_image=ip_adapter_image if ip_adapter_enable else None
    ).images[0]

    return image, seed

# @title Gradio
examples = [
    "homm3_spell_icon midivial sticker of a cartoon character of a man in a lab coat and glasses, old lady screaming and laughing",
    "homm3_spell_icon midivial sticker of a cartoon man with a mustache and a hat on, portrait bender from futurama, telegram sticker",
    "homm3_spell_icon midivial sticker of a cartoon character with a gun in his hand",
]

css = """
#col-container {
    margin: 0 auto;
    max-width: 640px;
}
"""

with gr.Blocks(css=css) as demo:
    with gr.Column(elem_id="col-container"):
        gr.Markdown(" # Text-to-Image Gradio Template")

        with gr.Row():
            # Dropdown to select the model from Hugging Face
            model_id = gr.Dropdown(
                label="Model",
                choices=MODEL_LIST,
                value=MODEL_LIST[0],  # Default model
            )

        with gr.Row():
            prompt = gr.Text(
                label="Prompt",
                show_label=False,
                max_lines=1,
                placeholder="Enter your prompt",
                container=False,
            )
            run_button = gr.Button("Run", scale=0, variant="primary")

        result = gr.Image(label="Result", show_label=False)

        with gr.Accordion("Advanced Settings", open=False):
            negative_prompt = gr.Text(
                label="Negative prompt",
                max_lines=1,
                placeholder="Enter a negative prompt",
            )

            seed = gr.Slider(
                label="Seed",
                minimum=0,
                maximum=MAX_SEED,
                step=1,
                value=42,  # Default seed
            )

            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=512,
                )
                height = gr.Slider(
                    label="Height",
                    minimum=256,
                    maximum=MAX_IMAGE_SIZE,
                    step=32,
                    value=512,
                )

            with gr.Row():
                guidance_scale = gr.Slider(
                    label="Guidance scale",
                    minimum=0.0,
                    maximum=20.0,
                    step=0.5,
                    value=7.0,
                )
                num_inference_steps = gr.Slider(
                    label="Number of inference steps",
                    minimum=1,
                    maximum=100,
                    step=1,
                    value=20,
                )

            # New slider for LoRA scale.
            lora_scale = gr.Slider(
                label="LoRA Scale",
                minimum=0.0,
                maximum=2.0,
                step=0.1,
                value=1.0,
                info="Adjust the influence of the LoRA weights",
            )

        # --- ControlNet Settings ---
        with gr.Accordion("ControlNet Settings", open=False):
            controlnet_enable = gr.Checkbox(
                label="Enable ControlNet",
                value=False
            )
            with gr.Group(visible=False) as controlnet_group:
                controlnet_mode = gr.Dropdown(
                    label="ControlNet Mode",
                    choices=list(CONTROLNET_MODES.keys()),
                    value=list(CONTROLNET_MODES.keys())[0],
                )
                controlnet_strength = gr.Slider(
                    label="ControlNet Conditioning Scale",
                    minimum=0.0,
                    maximum=1.0,
                    step=0.1,
                    value=0.7,
                )
                controlnet_image = gr.Image(
                    label="ControlNet Image",
                    type="pil"
                )

            def show_controlnet_options(enable):
                return {controlnet_group: gr.update(visible=enable)}

            controlnet_enable.change(
                fn=show_controlnet_options,
                inputs=controlnet_enable,
                outputs=controlnet_group,
            )

        # --- IP-adapter Settings ---
        with gr.Accordion("IP-adapter Settings", open=False):
            ip_adapter_enable = gr.Checkbox(
                label="Enable IP-adapter",
                value=False
            )

            with gr.Group(visible=False) as ip_adapter_group:
                ip_adapter_scale = gr.Slider(
                    label="IP-adapter Scale",
                    minimum=0.0,
                    maximum=2.0,
                    step=0.1,
                    value=1.0
                )
                ip_adapter_image = gr.Image(
                    label="IP-adapter Image",
                    type="pil"
                )

            # Show/hide IP-adapter parameters when checkbox is toggled
            def show_ip_adapter_options(enable):
                return {ip_adapter_group: gr.update(visible=enable)}

            ip_adapter_enable.change(
                fn=show_ip_adapter_options,
                inputs=ip_adapter_enable,
                outputs=ip_adapter_group,
            )

        gr.Examples(examples=examples, inputs=[prompt])
    gr.on(
        triggers=[run_button.click, prompt.submit],
        fn=infer,
        inputs=[
            model_id,
            prompt,
            negative_prompt,
            seed,
            randomize_seed,
            width,
            height,
            guidance_scale,
            num_inference_steps,
            lora_scale,
            controlnet_enable,
            controlnet_mode,
            controlnet_strength,
            controlnet_image,
            ip_adapter_enable,
            ip_adapter_scale,
            ip_adapter_image,
        ],
        outputs=[result, seed],
    )

# @title Run

if __name__ == "__main__":
    demo.launch(debug=True) # show errors in colab notebook