app.py

import spaces
import os
import numpy as np
import gradio as gr
import json
import torch
from diffusers import DiffusionPipeline
from transformers import CLIPTextModel, CLIPTokenizer, T5EncoderModel, T5TokenizerFast


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

# Use the 'waffles' environment variable as the access token
hf_token = os.getenv('waffles')

# Ensure the token is loaded correctly
if not hf_token:
    raise ValueError("Hugging Face API token not found. Please set the 'waffles' environment variable.")

# Load LoRAs from JSON file
with open('loras.json', 'r') as f:
    loras = json.load(f)

# Initialize the base model with authentication and specify the device
pipe = DiffusionPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16,
    token=hf_token
).to(device)

@spaces.GPU(duration=190)
MAX_SEED = 2**32-1

def run_lora(prompt, cfg_scale, steps, selected_repo, randomize_seed, seed, width, height, lora_scale):
    if not selected_repo:
        raise gr.Error("You must select a LoRA before proceeding.")

    selected_lora = next((lora for lora in loras if lora["repo"] == selected_repo), None)
    if not selected_lora:
        raise gr.Error("Selected LoRA not found.")

    lora_path = selected_lora["repo"]
    trigger_word = selected_lora["trigger_word"]

    # Load LoRA weights
    if "weights" in selected_lora:
        pipe.load_lora_weights(lora_path, weight_name=selected_lora["weights"])
    else:
        pipe.load_lora_weights(lora_path)
        
    # Set random seed for reproducibility
    if randomize_seed:
        seed = torch.randint(0, MAX_SEED, (1,)).item()

    # Generate image
    generator = torch.Generator(device=device).manual_seed(seed)
    image = pipe(
        prompt=f"{prompt} {trigger_word}",
        num_inference_steps=steps,
        guidance_scale=cfg_scale,
        width=width,
        height=height,
        generator=generator,
        joint_attention_kwargs={"scale": lora_scale},
    ).images[0]

    # Reset the model to CPU and unload LoRA weights to free up memory
    pipe.to("cpu")
    pipe.unload_lora_weights()

    return image, seed

with gr.Blocks() as app:
    with gr.Row():
        with gr.Column(scale=3):
            prompt = gr.Textbox(label="Prompt", lines=5, placeholder="Enter your prompt")
            lora_dropdown = gr.Dropdown(
                label="Select LoRA",
                choices=[lora["repo"] for lora in loras],
                value="XLabs-AI/flux-RealismLora",
            )
        with gr.Column(scale=1):
            generate_button = gr.Button("Generate", variant="primary")

    with gr.Row():
        result = gr.Image(label="Generated Image")
        seed = gr.Number(label="Seed", value=0, interactive=False)

    with gr.Accordion("Advanced Settings", open=False):
        cfg_scale = gr.Slider(label="CFG Scale", minimum=1, maximum=20, step=0.5, value=3.5)
        steps = gr.Slider(label="Steps", minimum=1, maximum=50, step=1, value=28)
        width = gr.Slider(label="Width", minimum=256, maximum=1536, step=64, value=1024)
        height = gr.Slider(label="Height", minimum=256, maximum=1536, step=64, value=1024)
        randomize_seed = gr.Checkbox(label="Randomize seed", value=True)
        lora_scale = gr.Slider(label="LoRA Scale", minimum=0, maximum=1, step=0.01, value=0.95)

    generate_button.click(
        run_lora,
        inputs=[prompt, cfg_scale, steps, lora_dropdown, randomize_seed, seed, width, height, lora_scale],
        outputs=[result, seed]
    )

app.launch()