Update app.py

app.py

@@ -1,5 +1,333 @@
 
+import gradio as gr
+import numpy as np
+import random
+import spaces
+import torch
+from diffusers import DiffusionPipeline, AutoencoderTiny, AutoencoderKL, StableDiffusionPipeline
+from live_preview_helpers import flux_pipe_call_that_returns_an_iterable_of_images
+from PIL import Image
+import io
+import zipfile
+from huggingface_hub import HfApi
+_HF_TOKEN = HfApi().token
+
+# --- Constants ---
+DTYPE = torch.bfloat16
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"
+MAX_SEED = np.iinfo(np.int32).max
+MAX_IMAGE_SIZE = 2048
+
+# --- Model Definitions ---
+MODELS = {
+    "FLUX.1-dev": {
+        "model_id": "black-forest-labs/FLUX.1-dev",
+        "vae_id": "madebyollin/taef1",
+        "pipeline_class": DiffusionPipeline,
+        "description": """
+            ## FLUX.1 [dev]
+            12B parameter rectified flow transformer guidance-distilled from [FLUX.1 [pro]](https://blackforestlabs.ai/).
+            - **License:** [Non-commercial license](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)
+            - **Blog:** [Read the announcement](https://blackforestlabs.ai/announcing-black-forest-labs/)
+            - **Model Card:** [View on Hugging Face](https://huggingface.co/black-forest-labs/FLUX.1-dev)
+            - **Developed by:** [Blackforest Labs](https://blackforestlabs.ai/)
+        """,
+        
+    },
+    "stable-diffusion-v1-5": {
+        "model_id": "runwayml/stable-diffusion-v1-5",
+        "pipeline_class": StableDiffusionPipeline,
+        "description": """
+            ## Stable Diffusion v1-5
+            Stable Diffusion v1-5 is a latent text-to-image diffusion model capable of generating photo-realistic images given any text input.
+            - **Model Card:** [View on Hugging Face](https://huggingface.co/runwayml/stable-diffusion-v1-5)
+            - **Developed by:** RunwayML
+            - **License:** [CreativeML Open RAIL-M License](https://huggingface.co/spaces/CompVis/stable-diffusion-license)
+        """,
+        "config": {
+            "requires_safety_checker":False
+        }
+    },
+    "deliberate-v3": {
+        "model_id": "XpucT/deliberate",
+        "pipeline_class": StableDiffusionPipeline,
+        "description": """
+            ## Deliberate V3
+            Deliberate V3 is a model merging designed for high quality image generation.
+            - **Model Card:** [View on Hugging Face](https://huggingface.co/XpucT/deliberate)
+            - **Developed by:** XpucT
+            - **License:** [CreativeML Open RAIL-M License](https://huggingface.co/spaces/CompVis/stable-diffusion-license)
+        """,
+        "config": {
+            "requires_safety_checker":False
+        }
+    },
+    "dreamshaper-8": {
+        "model_id": "Lykon/dreamshaper-8",
+        "pipeline_class": StableDiffusionPipeline,
+        "description": """
+            ## DreamShaper 8
+            DreamShaper 8 is another iteration of the fine-tuned stable diffusion model that is capable of producing high-quality and detailed images.
+            - **Model Card:** [View on Hugging Face](https://huggingface.co/Lykon/dreamshaper-8)
+            - **Developed by:** Lykon
+            - **License:** [CreativeML Open RAIL-M License](https://huggingface.co/spaces/CompVis/stable-diffusion-license)
+        """,
+        "config": {
+            "requires_safety_checker":False
+        }
+    },
+    # Add other models here in the same format
+}
 
+
+MODELS  = {
+    "FLUX.1-schnell": {
+        "pipeline_class": FluxPipeline,
+        "model_id": "black-forest-labs/FLUX.1-schnell",
+        "config": {"torch_dtype": torch.bfloat16},
+        "description": "**FLUX.1-schnell** is a fast and efficient model designed for quick image generation. It excels at producing high-quality images rapidly, making it ideal for applications where speed is crucial. However, its rapid generation may slightly compromise on the level of detail compared to slower, more meticulous models.",
+    },
+}
+# --- Function to pre-download models ---
+def download_all_models():
+    print("Downloading all models...")
+    for model_key, config in MODELS.items():
+        try:
+            pipeline_class = config["pipeline_class"]
+            model_id = config["model_id"]
+            # Attempt to download the pipeline without loading it into memory
+            pipeline_class.download(model_id, token=_HF_TOKEN, **config.get("config", {}))
+
+            print(f"Model '{model_key}' downloaded successfully.")
+        except Exception as e:
+            print(f"Error downloading model '{model_key}': {e}")
+    print("Model download process complete.")
+
+# --- Function to clear GPU memory ---
+def clear_gpu_memory():
+    if DEVICE == "cuda":
+        with torch.no_grad():
+            torch.cuda.empty_cache()
+
+# --- Function to load models and setup pipeline ---
+def load_models(model_key):
+    clear_gpu_memory()
+    model_info = MODELS[model_key]
+    pipeline_class = model_info["pipeline_class"]
+
+    if "vae_id" in model_info:
+      vae = AutoencoderTiny.from_pretrained(model_info["vae_id"], torch_dtype=DTYPE).to(DEVICE)
+      good_vae = AutoencoderKL.from_pretrained(model_info["model_id"], subfolder="vae", torch_dtype=DTYPE).to(DEVICE)
+      pipe = pipeline_class.from_pretrained(model_info["model_id"], torch_dtype=DTYPE, vae=vae, **model_info.get("config", {})).to(DEVICE)
+      pipe.flux_pipe_call_that_returns_an_iterable_of_images = flux_pipe_call_that_returns_an_iterable_of_images.__get__(pipe)
+      return pipe, good_vae
+
+    else:
+      pipe = pipeline_class.from_pretrained(model_info["model_id"], torch_dtype=DTYPE, **model_info.get("config", {})).to(DEVICE)
+      return pipe, None
+
+# --- Initial model load ---
+current_model_key = "FLUX.1-dev"  # Start with FLUX.1-dev
+pipe, good_vae = load_models(current_model_key)
+
+# --- Inference function ---
+@spaces.GPU(duration=75)
+def infer(model_key, prompt, seed=42, randomize_seed=False, width=1024, height=1024, guidance_scale=3.5, num_inference_steps=28, progress=gr.Progress(track_tqdm=True)):
+    global pipe, good_vae, current_model_key
+    if model_key != current_model_key:
+        pipe, good_vae = load_models(model_key)
+        current_model_key = model_key
+
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
+    generator = torch.Generator().manual_seed(seed)
+
+    images = []
+
+    if current_model_key == "FLUX.1-dev":
+        for img in pipe.flux_pipe_call_that_returns_an_iterable_of_images(
+            prompt=prompt,
+            guidance_scale=guidance_scale,
+            num_inference_steps=num_inference_steps,
+            width=width,
+            height=height,
+            generator=generator,
+            output_type="pil",
+            good_vae=good_vae,
+        ):
+            images.append(img)
+            yield img, seed, None
+    else:
+        result  = pipe(
+            prompt=prompt,
+            guidance_scale=guidance_scale,
+            num_inference_steps=num_inference_steps,
+            width=width,
+            height=height,
+            generator=generator,
+            
+        )
+        images.extend(result.images)
+
+        for img in result.images:
+          yield img, seed, None
+
+    if images:
+        zip_buffer = io.BytesIO()
+        with zipfile.ZipFile(zip_buffer, "w") as zf:
+            for i, img in enumerate(images):
+                img_buffer = io.BytesIO()
+                img.save(img_buffer, format="PNG")
+                zf.writestr(f"image_{i}.png", img_buffer.getvalue())
+        yield images[-1], seed, zip_buffer
+    else:
+        yield None, seed, None
+
+# --- Example prompts ---
+examples = [
+    "a tiny astronaut hatching from an egg on the moon",
+    "a cat holding a sign that says hello world",
+    "an anime illustration of a wiener schnitzel",
+]
+
+# --- CSS for styling ---
+css = """
+#col-container {
+    margin-left: auto;
+    margin-right: auto;
+    text-align: center;
+}
+.text-center {
+    text-align: center;
+}
+.title {
+    font-size: 1.5rem;
+    font-weight: bold;
+    margin-bottom: 1rem;
+}
+.footer {
+    text-align: center;
+    margin-top: 1rem;
+}
+.description-text {
+    text-align: left;
+    margin-bottom: 1rem;
+    
+}
+"""
+
+# --- Gradio Interface ---
+with gr.Blocks(css=css) as demo:
+    with gr.Column(elem_id="col-container"):
+        gr.Markdown(
+            """
+            <div class="title">
+            🖼️ AI Image Generator 🖼️
+            </div>
+            <div class="text-center">
+            Choose a model and generate stunning images with AI!
+            </div>
+            """,
+        )
+        
+        with gr.Tab("Generator"):
+            with gr.Row():
+                model_selector = gr.Dropdown(
+                    label="Select Model",
+                    choices=list(MODELS.keys()),
+                    value=current_model_key,
+                )
+
+            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)
+            result = gr.Image(label="Result", show_label=False, elem_id="result-image")
+            download_button = gr.Button("Download Results")
+            with gr.Accordion("Advanced Settings", open=False):
+                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.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=1024,
+                    )
+                with gr.Row():
+                    guidance_scale = gr.Slider(
+                        label="Guidance Scale",
+                        minimum=1,
+                        maximum=15,
+                        step=0.1,
+                        value=3.5,
+                    )
+                    num_inference_steps = gr.Slider(
+                        label="Number of inference steps",
+                        minimum=1,
+                        maximum=50,
+                        step=1,
+                        value=28,
+                    )
+            gr.Examples(
+                examples=examples,
+                fn=infer,
+                inputs=[model_selector, prompt],
+                outputs=[result, seed, download_button],
+                cache_examples="lazy",
+            )
+        with gr.Tab("Model Descriptions"):
+            for model_key, model_info in MODELS.items():
+                with gr.Accordion(model_key, open=False):
+                    gr.Markdown(model_info["description"], elem_classes="description-text")
+
+        gr.Markdown(
+            """
+            <div class="footer">
+            <p>
+            ⚡ Powered by <a href="https://www.gradio.app/" target="_blank">Gradio</a> and <a href="https://huggingface.co/spaces" target="_blank">🤗 Spaces</a>.
+            </p>
+            </div>
+            """,
+        )
+
+    # --- Event handlers ---
+    inputs = [model_selector, prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps]
+    outputs = [result, seed, download_button]
+    gr.on(
+        triggers=[run_button.click, prompt.submit],
+        fn=infer,
+        inputs=inputs,
+        outputs=outputs,
+    )
+    download_event = download_button.click(lambda x: x, inputs=download_button, outputs=download_button, queue=False)
+
+# --- Pre-download all models at startup ---
+download_all_models()
+
+# --- Launch the demo ---
+demo.queue().launch(debug=True)
+
+'''
 import os
 import torch
 import gradio as gr
@@ -221,7 +549,7 @@ if __name__ == "__main__":
     download_all_models()      
     interface.launch(debug=True)
 
-
+'''