Create app-backup.py

app-backup.py

@@ -0,0 +1,700 @@
+import random
+import os
+import uuid
+from datetime import datetime
+import gradio as gr
+import numpy as np
+import spaces
+import torch
+from diffusers import DiffusionPipeline
+from PIL import Image
+
+# Apply more comprehensive patches to Gradio's utility functions
+import gradio_client.utils
+import types
+
+# Patch 1: Fix the _json_schema_to_python_type function
+original_json_schema = gradio_client.utils._json_schema_to_python_type
+
+def patched_json_schema(schema, defs=None):
+    # Handle boolean values directly
+    if isinstance(schema, bool):
+        return "bool"
+    
+    # Handle cases where 'additionalProperties' is a boolean
+    try:
+        if "additionalProperties" in schema and isinstance(schema["additionalProperties"], bool):
+            schema["additionalProperties"] = {"type": "any"}
+    except (TypeError, KeyError):
+        pass
+    
+    # Call the original function
+    try:
+        return original_json_schema(schema, defs)
+    except Exception as e:
+        # Fallback to a safe value when the schema can't be parsed
+        return "any"
+
+# Replace the original function with our patched version
+gradio_client.utils._json_schema_to_python_type = patched_json_schema
+
+# Create permanent storage directory
+SAVE_DIR = "saved_images"  # Gradio will handle the persistence
+if not os.path.exists(SAVE_DIR):
+    os.makedirs(SAVE_DIR, exist_ok=True)
+
+# Safe settings for model loading
+device = "cuda" if torch.cuda.is_available() else "cpu"
+repo_id = "black-forest-labs/FLUX.1-dev"
+adapter_id = "openfree/flux-chatgpt-ghibli-lora"
+
+def load_model_with_retry(max_retries=5):
+    for attempt in range(max_retries):
+        try:
+            print(f"Loading model attempt {attempt+1}/{max_retries}...")
+            pipeline = DiffusionPipeline.from_pretrained(
+                repo_id, 
+                torch_dtype=torch.bfloat16,
+                use_safetensors=True,
+                resume_download=True
+            )
+            print("Model loaded successfully, loading LoRA weights...")
+            pipeline.load_lora_weights(adapter_id)
+            pipeline = pipeline.to(device)
+            print("Pipeline ready!")
+            return pipeline
+        except Exception as e:
+            if attempt < max_retries - 1:
+                wait_time = 10 * (attempt + 1)
+                print(f"Error loading model: {e}. Retrying in {wait_time} seconds...")
+                import time
+                time.sleep(wait_time)
+            else:
+                raise Exception(f"Failed to load model after {max_retries} attempts: {e}")
+
+# Load the model
+pipeline = load_model_with_retry()
+
+MAX_SEED = np.iinfo(np.int32).max
+MAX_IMAGE_SIZE = 1024
+
+def save_generated_image(image, prompt):
+    # Generate unique filename with timestamp
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    unique_id = str(uuid.uuid4())[:8]
+    filename = f"{timestamp}_{unique_id}.png"
+    filepath = os.path.join(SAVE_DIR, filename)
+    
+    # Save the image
+    image.save(filepath)
+    
+    # Save metadata
+    metadata_file = os.path.join(SAVE_DIR, "metadata.txt")
+    with open(metadata_file, "a", encoding="utf-8") as f:
+        f.write(f"{filename}|{prompt}|{timestamp}\n")
+    
+    return filepath
+
+def load_generated_images():
+    if not os.path.exists(SAVE_DIR):
+        return []
+    
+    # Load all images from the directory
+    image_files = [
+        os.path.join(SAVE_DIR, f)
+        for f in os.listdir(SAVE_DIR)
+        if f.endswith(('.png', '.jpg', '.jpeg', '.webp'))
+    ]
+    # Sort by creation time (newest first)
+    image_files.sort(key=lambda x: os.path.getctime(x), reverse=True)
+    return image_files
+
+@spaces.GPU(duration=120)
+def inference(
+    prompt: str,
+    seed: int,
+    randomize_seed: bool,
+    width: int,
+    height: int,
+    guidance_scale: float,
+    num_inference_steps: int,
+    lora_scale: float,
+    progress: gr.Progress = gr.Progress(track_tqdm=True),
+):
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
+    generator = torch.Generator(device=device).manual_seed(seed)
+    
+    try:
+        image = pipeline(
+            prompt=prompt,
+            guidance_scale=guidance_scale,
+            num_inference_steps=num_inference_steps,
+            width=width,
+            height=height,
+            generator=generator,
+            joint_attention_kwargs={"scale": lora_scale},
+        ).images[0]
+        
+        # Save the generated image
+        filepath = save_generated_image(image, prompt)
+        
+        # Return the image, seed, and updated gallery
+        return image, seed, load_generated_images()
+    except Exception as e:
+        print(f"Error during inference: {e}")
+        error_img = Image.new('RGB', (width, height), color='red')
+        return error_img, seed, load_generated_images()
+
+examples = [
+    "Ghibli style futuristic stormtrooper with glossy white armor and a sleek helmet, standing heroically on a lush alien planet, vibrant flowers blooming around, soft sunlight illuminating the scene, a gentle breeze rustling the leaves. The armor reflects the pink and purple hues of the alien sunset, creating an ethereal glow around the figure. [trigger]",
+    "Ghibli style young mechanic girl in a floating workshop, surrounded by hovering tools and glowing mechanical parts, her blue overalls covered in oil stains, tinkering with a semi-transparent robot companion. Magical sparks fly as she works, while floating islands with waterfalls drift past her open workshop window. [trigger]",
+    "Ghibli style ancient forest guardian robot, covered in moss and flowering vines, sitting peacefully in a crystal-clear lake. Its gentle eyes glow with soft blue light, while bioluminescent dragonflies dance around its weathered metal frame. Ancient tech symbols on its surface pulse with a gentle rhythm. [trigger]",
+    "Ghibli style sky whale transport ship, its metallic skin adorned with traditional Japanese patterns, gliding through cotton candy clouds at sunrise. Small floating gardens hang from its sides, where workers in futuristic kimonos tend to glowing plants. Rainbow auroras shimmer in the background. [trigger]",
+    "Ghibli style cyber-shrine maiden with flowing holographic robes, performing a ritual dance among floating lanterns and digital cherry blossoms. Her traditional headdress emits soft light patterns, while spirit-like AI constructs swirl around her in elegant patterns. The scene is set in a modern shrine with both ancient wood and sleek chrome elements. [trigger]",
+    "Ghibli style robot farmer tending to floating rice paddies in the sky, wearing a traditional straw hat with advanced sensors. Its gentle movements create ripples in the water as it plants glowing rice seedlings. Flying fish leap between the terraced fields, leaving trails of sparkles in their wake, while future Tokyo's spires gleam in the distance. [trigger]"
+]
+
+# Enhanced CSS for a more visually refined UI
+css = """
+:root {
+    --primary-color: #6a92cc;
+    --primary-hover: #557ab8;
+    --secondary-color: #f4c062;
+    --background-color: #f7f9fc;
+    --panel-background: #ffffff;
+    --text-color: #333333;
+    --border-radius: 12px;
+    --shadow: 0 4px 12px rgba(0,0,0,0.08);
+    --font-main: 'Poppins', -apple-system, BlinkMacSystemFont, sans-serif;
+}
+
+body {
+    background-color: var(--background-color);
+    font-family: var(--font-main);
+}
+
+.gradio-container {
+    margin: 0 auto;
+    max-width: 1200px !important;
+}
+
+.main-header {
+    text-align: center;
+    padding: 2rem 1rem 1rem;
+    background: linear-gradient(90deg, #6a92cc 0%, #8f7fc8 100%);
+    color: white;
+    margin-bottom: 2rem;
+    border-radius: var(--border-radius);
+    box-shadow: var(--shadow);
+}
+
+.main-header h1 {
+    font-size: 2.5rem;
+    margin-bottom: 0.5rem;
+    font-weight: 700;
+    text-shadow: 0 2px 4px rgba(0,0,0,0.2);
+}
+
+.main-header p {
+    font-size: 1rem;
+    margin-bottom: 0.5rem;
+    opacity: 0.9;
+}
+
+.main-header a {
+    color: var(--secondary-color);
+    text-decoration: none;
+    font-weight: 600;
+    transition: all 0.2s ease;
+}
+
+.main-header a:hover {
+    text-decoration: underline;
+    opacity: 0.9;
+}
+
+.container {
+    background-color: var(--panel-background);
+    padding: 1.5rem;
+    border-radius: var(--border-radius);
+    box-shadow: var(--shadow);
+    margin-bottom: 1.5rem;
+}
+
+button.primary {
+    background: var(--primary-color) !important;
+    border: none !important;
+    color: white !important;
+    padding: 10px 20px !important;
+    border-radius: 8px !important;
+    font-weight: 600 !important;
+    box-shadow: 0 2px 5px rgba(0,0,0,0.1) !important;
+    transition: all 0.2s ease !important;
+}
+
+button.primary:hover {
+    background: var(--primary-hover) !important;
+    transform: translateY(-2px) !important;
+    box-shadow: 0 4px 8px rgba(0,0,0,0.15) !important;
+}
+
+button.secondary {
+    background: white !important;
+    border: 1px solid #ddd !important;
+    color: var(--text-color) !important;
+    padding: 10px 20px !important;
+    border-radius: 8px !important;
+    font-weight: 500 !important;
+    box-shadow: 0 2px 5px rgba(0,0,0,0.05) !important;
+    transition: all 0.2s ease !important;
+}
+
+button.secondary:hover {
+    background: #f5f5f5 !important;
+    transform: translateY(-2px) !important;
+}
+
+.gr-box {
+    border-radius: var(--border-radius) !important;
+    border: 1px solid #e0e0e0 !important;
+}
+
+.gr-panel {
+    border-radius: var(--border-radius) !important;
+}
+
+.gr-input {
+    border-radius: 8px !important;
+    border: 1px solid #ddd !important;
+    padding: 12px !important;
+}
+
+.gr-form {
+    border-radius: var(--border-radius) !important;
+    background-color: var(--panel-background) !important;
+}
+
+.gr-accordion {
+    border-radius: var(--border-radius) !important;
+    overflow: hidden !important;
+}
+
+.gr-button {
+    border-radius: 8px !important;
+}
+
+.gallery-item {
+    border-radius: var(--border-radius) !important;
+    transition: all 0.3s ease !important;
+}
+
+.gallery-item:hover {
+    transform: scale(1.02) !important;
+    box-shadow: 0 6px 15px rgba(0,0,0,0.1) !important;
+}
+
+.tabs {
+    border-radius: var(--border-radius) !important;
+    overflow: hidden !important;
+}
+
+footer {
+    display: none !important;
+}
+
+.settings-accordion legend span {
+    font-weight: 600 !important;
+}
+
+.example-prompt {
+    font-size: 0.9rem;
+    color: #555;
+    padding: 8px;
+    background: #f5f7fa;
+    border-radius: 6px;
+    border-left: 3px solid var(--primary-color);
+    margin-bottom: 8px;
+    cursor: pointer;
+    transition: all 0.2s;
+}
+
+.example-prompt:hover {
+    background: #eef2f8;
+}
+
+.status-generating {
+    color: #ffa200;
+    font-weight: 500;
+    display: flex;
+    align-items: center;
+    gap: 8px;
+}
+
+.status-generating::before {
+    content: "";
+    display: inline-block;
+    width: 12px;
+    height: 12px;
+    border-radius: 50%;
+    background-color: #ffa200;
+    animation: pulse 1.5s infinite;
+}
+
+.status-complete {
+    color: #00c853;
+    font-weight: 500;
+    display: flex;
+    align-items: center;
+    gap: 8px;
+}
+
+.status-complete::before {
+    content: "";
+    display: inline-block;
+    width: 12px;
+    height: 12px;
+    border-radius: 50%;
+    background-color: #00c853;
+}
+
+@keyframes pulse {
+    0% {
+        opacity: 0.6;
+    }
+    50% {
+        opacity: 1;
+    }
+    100% {
+        opacity: 0.6;
+    }
+}
+
+.gr-accordion-title {
+    font-weight: 600 !important;
+    color: var(--text-color) !important;
+}
+
+.tabs button {
+    font-weight: 500 !important;
+    padding: 10px 16px !important;
+}
+
+.tabs button.selected {
+    font-weight: 600 !important;
+    color: var(--primary-color) !important;
+    background: rgba(106, 146, 204, 0.1) !important;
+}
+
+.gr-slider-container {
+    padding: 10px 0 !important;
+}
+
+.gr-prose h3 {
+    font-weight: 600 !important;
+    color: var(--primary-color) !important;
+    margin-bottom: 1rem !important;
+}
+"""
+
+#######################
+# SECOND TAB CODE SETUP
+#######################
+import subprocess
+subprocess.run(
+    'pip install flash-attn --no-build-isolation',
+    env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"},
+    shell=True
+)
+
+from transformers import AutoProcessor, AutoModelForCausalLM
+
+# Load your Florence-2 models and processors
+models = {
+    'gokaygokay/Florence-2-Flux-Large': AutoModelForCausalLM.from_pretrained('gokaygokay/Florence-2-Flux-Large', trust_remote_code=True).eval(),
+    'gokaygokay/Florence-2-Flux': AutoModelForCausalLM.from_pretrained('gokaygokay/Florence-2-Flux', trust_remote_code=True).eval(),
+}
+
+processors = {
+    'gokaygokay/Florence-2-Flux-Large': AutoProcessor.from_pretrained('gokaygokay/Florence-2-Flux-Large', trust_remote_code=True),
+    'gokaygokay/Florence-2-Flux': AutoProcessor.from_pretrained('gokaygokay/Florence-2-Flux', trust_remote_code=True),
+}
+
+title = """<h1 align="center">Florence-2 Captioner for Flux Prompts</h1>
+<p><center>
+<a href="https://huggingface.co/gokaygokay/Florence-2-Flux-Large" target="_blank">[Florence-2 Flux Large]</a>
+<a href="https://huggingface.co/gokaygokay/Florence-2-Flux" target="_blank">[Florence-2 Flux Base]</a>
+</center></p>
+"""
+
+@spaces.GPU
+def run_example(image, model_name='gokaygokay/Florence-2-Flux-Large'):
+    from PIL import Image as PILImage
+    image = PILImage.fromarray(image)
+    task_prompt = "<DESCRIPTION>"
+    prompt = task_prompt + "Describe this image in great detail."
+
+    if image.mode != "RGB":
+        image = image.convert("RGB")
+
+    model = models[model_name]
+    processor = processors[model_name]
+
+    inputs = processor(text=prompt, images=image, return_tensors="pt")
+    generated_ids = model.generate(
+        input_ids=inputs["input_ids"],
+        pixel_values=inputs["pixel_values"],
+        max_new_tokens=1024,
+        num_beams=3,
+        repetition_penalty=1.10,
+    )
+    generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
+    parsed_answer = processor.post_process_generation(generated_text, task=task_prompt, image_size=(image.width, image.height))
+    return parsed_answer["<DESCRIPTION>"]
+
+
+####################################################
+# COMBINED DEMO WITH TWO TABS
+####################################################
+with gr.Blocks(css=css, analytics_enabled=False, theme="soft") as demo:
+    with gr.Column():
+        # Custom header (Tab 1)
+        gr.HTML('''
+        <div class="main-header">
+            <h1>✨ FLUX Ghibli LoRA Generator ✨</h1>
+            <p>Community: <a href="https://discord.gg/openfreeai" target="_blank">https://discord.gg/openfreeai</a></p>
+        </div>
+        ''')
+        
+        with gr.Row():
+            with gr.Column(scale=3):
+                with gr.Group(elem_classes="container"):
+                    prompt = gr.Textbox(
+                        label="Enter your imagination",
+                        placeholder="Describe your Ghibli-style image here...",
+                        lines=3
+                    )
+                    
+                    with gr.Row():
+                        run_button = gr.Button("✨ Generate Image", elem_classes="primary")
+                        clear_button = gr.Button("Clear", elem_classes="secondary")
+                    
+                    with gr.Accordion("Advanced Settings", open=False, elem_classes="settings-accordion"):
+                        with gr.Row():
+                            seed = gr.Slider(
+                                label="Seed",
+                                minimum=0,
+                                maximum=MAX_SEED,
+                                step=1,
+                                value=42,
+                            )
+                            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=768,
+                            )
+
+                        with gr.Row():
+                            guidance_scale = gr.Slider(
+                                label="Guidance scale",
+                                minimum=0.0,
+                                maximum=10.0,
+                                step=0.1,
+                                value=3.5,
+                            )
+                            
+                        with gr.Row():
+                            num_inference_steps = gr.Slider(
+                                label="Steps",
+                                minimum=1,
+                                maximum=50,
+                                step=1,
+                                value=30,
+                            )
+                            lora_scale = gr.Slider(
+                                label="LoRA scale",
+                                minimum=0.0,
+                                maximum=1.0,
+                                step=0.1,
+                                value=1.0,
+                            )
+                
+                with gr.Group(elem_classes="container"):
+                    gr.Markdown("### ✨ Example Prompts")
+                    # Create HTML for examples manually
+                    examples_html = '\n'.join([
+                        f'<div class="example-prompt">{example}</div>' 
+                        for example in examples
+                    ])
+                    example_container = gr.HTML(examples_html)
+            
+            with gr.Column(scale=4):
+                with gr.Group(elem_classes="container"):
+                    # Image result container with status indicator
+                    with gr.Group():
+                        generation_status = gr.HTML('<div class="status-complete">Ready to generate</div>')
+                        result = gr.Image(label="Generated Image", elem_id="result-image")
+                        seed_text = gr.Number(label="Used Seed", value=42)
+
+    # TABS
+    with gr.Tabs(elem_classes="tabs") as tabs:
+        with gr.TabItem("Gallery"):
+            with gr.Group(elem_classes="container"):
+                gallery_header = gr.Markdown("### 🖼️ Your Generated Masterpieces")
+                with gr.Row():
+                    refresh_btn = gr.Button("🔄 Refresh Gallery", elem_classes="secondary")
+                generated_gallery = gr.Gallery(
+                    label="Generated Images",
+                    columns=3,
+                    value=load_generated_images(),
+                    height="500px",
+                    elem_classes="gallery-item"
+                )
+
+        #####################################
+        # SECOND TAB: FLORENCE-2 CAPTIONER
+        #####################################
+        with gr.TabItem("Florence-2 Captioner"):
+            # Use your custom HTML title
+            gr.HTML(title)
+
+            with gr.Row():
+                with gr.Column():
+                    input_img = gr.Image(label="Input Picture")
+                    model_selector = gr.Dropdown(
+                        choices=list(models.keys()),
+                        label="Model",
+                        value='gokaygokay/Florence-2-Flux-Large'
+                    )
+                    submit_btn = gr.Button(value="Submit")
+                with gr.Column():
+                    output_text = gr.Textbox(label="Output Text")
+
+            gr.Examples(
+                [["image1.jpg"], 
+                 ["image2.jpg"], 
+                 ["image3.png"], 
+                 ["image5.jpg"]],
+                inputs=[input_img, model_selector],
+                outputs=[output_text],
+                fn=run_example,
+                label='Try captioning on below examples',
+                cache_examples=True
+            )
+
+            submit_btn.click(run_example, [input_img, model_selector], [output_text])
+    # END TABS
+
+    # ----- EVENT HANDLERS for Tab 1 -----
+    def refresh_gallery():
+        return load_generated_images()
+
+    def clear_output():
+        return "", gr.update(value=None), seed, '<div class="status-complete">Ready to generate</div>'
+    
+    def before_generate():
+        return '<div class="status-generating">Generating image...</div>'
+    
+    def after_generate(image, seed, gallery):
+        return image, seed, gallery, '<div class="status-complete">Generation complete!</div>'
+
+    refresh_btn.click(
+        fn=refresh_gallery,
+        inputs=None,
+        outputs=generated_gallery,
+    )
+    
+    clear_button.click(
+        fn=clear_output,
+        inputs=None,
+        outputs=[prompt, result, seed_text, generation_status]
+    )
+    
+    run_button.click(
+        fn=before_generate,
+        inputs=None,
+        outputs=generation_status,
+    ).then(
+        fn=inference,
+        inputs=[
+            prompt,
+            seed,
+            randomize_seed,
+            width,
+            height,
+            guidance_scale,
+            num_inference_steps,
+            lora_scale,
+        ],
+        outputs=[result, seed_text, generated_gallery],
+    ).then(
+        fn=after_generate,
+        inputs=[result, seed_text, generated_gallery],
+        outputs=[result, seed_text, generated_gallery, generation_status],
+    )
+    
+    prompt.submit(
+        fn=before_generate,
+        inputs=None,
+        outputs=generation_status,
+    ).then(
+        fn=inference,
+        inputs=[
+            prompt,
+            seed,
+            randomize_seed,
+            width,
+            height,
+            guidance_scale,
+            num_inference_steps,
+            lora_scale,
+        ],
+        outputs=[result, seed_text, generated_gallery],
+    ).then(
+        fn=after_generate,
+        inputs=[result, seed_text, generated_gallery],
+        outputs=[result, seed_text, generated_gallery, generation_status],
+    )
+    
+    # Custom JavaScript for handling example prompts
+    gr.HTML("""
+    <script>
+    document.addEventListener('DOMContentLoaded', function() {
+        // Add click handlers to example prompts
+        setTimeout(() => {
+            const examples = document.querySelectorAll('.example-prompt');
+            const promptInput = document.querySelector('textarea');
+            
+            examples.forEach(example => {
+                example.addEventListener('click', function() {
+                    promptInput.value = this.textContent.trim();
+                    // Trigger input event to update Gradio's state
+                    const event = new Event('input', { bubbles: true });
+                    promptInput.dispatchEvent(event);
+                });
+            });
+        }, 1000); // Small delay to ensure elements are loaded
+    });
+    </script>
+    """)
+
+# Launch with fallback options
+try:
+    demo.queue(concurrency_count=1, max_size=20)
+    demo.launch(debug=True, show_api=False)
+except Exception as e:
+    print(f"Error during launch: {e}")
+    print("Trying alternative launch configuration...")
+    demo.launch(debug=True, show_api=False, share=False)