Update app.py

app.py

@@ -1,83 +1,157 @@
 import gradio as gr
 import numpy as np
 import random
-import os  # Import os to access secrets
-from diffusers import DiffusionPipeline
+import spaces
 import torch
+import time
+from diffusers import DiffusionPipeline, AutoencoderTiny
+from custom_pipeline import FLUXPipelineWithIntermediateOutputs
 
-# Set device
-device = "cuda" if torch.cuda.is_available() else "cpu"
-
-# Use Hugging Face secret from your space environment
-access_token = os.getenv('HUGGINGFACE_TOKEN')
-
-# Replace with the correct model repo ID
-model_repo_id = "KingNish/Realtime-FLUX"
-
-# Set the torch dtype based on availability of CUDA
-if torch.cuda.is_available():
-    torch_dtype = torch.float16
-else:
-    torch_dtype = torch.float32
-
-# Load the pipeline with the access token
-pipe = DiffusionPipeline.from_pretrained(model_repo_id, torch_dtype=torch_dtype, use_auth_token=access_token).to(device)
-
+# Constants
 MAX_SEED = np.iinfo(np.int32).max
-MAX_IMAGE_SIZE = 1024
+MAX_IMAGE_SIZE = 2048
+DEFAULT_WIDTH = 1024
+DEFAULT_HEIGHT = 1024
+DEFAULT_INFERENCE_STEPS = 1
 
-def infer(prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, progress=gr.Progress(track_tqdm=True)):
+# Device and model setup for CPU
+dtype = torch.float32  # Change to float32 since it's more suitable for CPU
+pipe = FLUXPipelineWithIntermediateOutputs.from_pretrained(
+    "black-forest-labs/FLUX.1-schnell", torch_dtype=dtype
+).to("cpu")  # Running on CPU
+pipe.vae = AutoencoderTiny.from_pretrained("madebyollin/taef1", torch_dtype=torch.float32).to("cpu")
+torch.cuda.empty_cache()  # No need for this line in CPU setup
+
+# Inference function
+@spaces.CPU(duration=25)  # Adjusted to CPU decorator
+def generate_image(prompt, seed=24, width=DEFAULT_WIDTH, height=DEFAULT_HEIGHT, randomize_seed=False, num_inference_steps=2, progress=gr.Progress(track_tqdm=True)):
     if randomize_seed:
         seed = random.randint(0, MAX_SEED)
-    
-    generator = torch.Generator().manual_seed(seed)
-    
-    image = pipe(
-        prompt=prompt, 
-        negative_prompt=negative_prompt,
-        guidance_scale=guidance_scale, 
-        num_inference_steps=num_inference_steps, 
-        width=width, 
+    generator = torch.Generator().manual_seed(int(float(seed)))
+
+    start_time = time.time()
+
+    # Only generate the last image in the sequence
+    img = pipe.generate_images(
+        prompt=prompt,
+        width=width,
         height=height,
+        num_inference_steps=num_inference_steps,
         generator=generator
-    ).images[0] 
-    
-    return image, seed
+    )
+    latency = f"Latency: {(time.time()-start_time):.2f} seconds"
+    return img, seed, latency
 
+# Example prompts
 examples = [
-    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
-    "An astronaut riding a green horse",
-    "A delicious ceviche cheesecake slice",
+    "a tiny astronaut hatching from an egg on the moon",
+    "a cute white cat holding a sign that says hello world",
+    "an anime illustration of Steve Jobs",
+    "Create image of Modern house in minecraft style",
+    "photo of a woman on the beach, shot from above. She is facing the sea, while wearing a white dress. She has long blonde hair",
+    "Selfie photo of a wizard with long beard and purple robes, he is apparently in the middle of Tokyo. Probably taken from a phone.",
+    "Photo of a young woman with long, wavy brown hair tied in a bun and glasses. She has a fair complexion and is wearing subtle makeup, emphasizing her eyes and lips. She is dressed in a black top. The background appears to be an urban setting with a building facade, and the sunlight casts a warm glow on her face.",
 ]
 
-css = """
-#col-container {
-    margin: 0 auto;
-    max-width: 640px;
-}
-"""
+# --- Gradio UI ---
+with gr.Blocks() as demo:
+    with gr.Column(elem_id="app-container"):
+        gr.Markdown("# FLUX Image Generator Demo")
+        gr.Markdown("Generate stunning images in real-time with Modified Flux.Schnell pipeline.")
+        #gr.Markdown("<span style='color: red;'>Note: Sometimes it stucks or stops generating images (I don't know why). In that situation just refresh the site.</span>")
 
-with gr.Blocks(css=css) as demo:
-    with gr.Column(elem_id="col-container"):
-        gr.Markdown("# Text-to-Image Gradio Template")
         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)
-
-        with gr.Accordion("Advanced Settings", open=False):
-            negative_prompt = gr.Text(label="Negative prompt", max_lines=1, placeholder="Enter a negative prompt", visible=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)
-            guidance_scale = gr.Slider(label="Guidance scale", minimum=0.0, maximum=10.0, step=0.1, value=7.5)
-            num_inference_steps = gr.Slider(label="Number of inference steps", minimum=1, maximum=50, step=1, value=25)
-        
-        gr.Examples(examples=examples, inputs=[prompt])
-
-    gr.on(run_button.click, infer, inputs=[prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps], outputs=[result, seed])
-
-demo.queue().launch()
+            with gr.Column(scale=2.5):
+                result = gr.Image(label="Generated Image", show_label=False, interactive=False)
+            with gr.Column(scale=1):
+                prompt = gr.Text(
+                    label="Prompt",
+                    placeholder="Describe the image you want to generate...",
+                    lines=3,
+                    show_label=False,
+                    container=False,
+                )
+                generateBtn = gr.Button("🖼️ Generate Image")
+                enhanceBtn = gr.Button("🚀 Enhance Image")
+
+                with gr.Column("Advanced Options"):
+                    with gr.Row():
+                        realtime = gr.Checkbox(label="Realtime Toggler", info="If TRUE then uses more CPU resources but creates the image in real-time.", value=False)
+                        latency = gr.Text(label="Latency")
+                    with gr.Row():
+                        seed = gr.Number(label="Seed", 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=DEFAULT_WIDTH)
+                        height = gr.Slider(label="Height", minimum=256, maximum=MAX_IMAGE_SIZE, step=32, value=DEFAULT_HEIGHT)
+                        num_inference_steps = gr.Slider(label="Inference Steps", minimum=1, maximum=4, step=1, value=DEFAULT_INFERENCE_STEPS)
+
+        with gr.Row():
+            gr.Markdown("### 🌟 Inspiration Gallery")
+        with gr.Row():
+            gr.Examples(
+                examples=examples,
+                fn=generate_image,
+                inputs=[prompt],
+                outputs=[result, seed, latency],
+                cache_examples="lazy" 
+            )
+
+    enhanceBtn.click(
+        fn=generate_image,
+        inputs=[prompt, seed, width, height],
+        outputs=[result, seed, latency],
+        show_progress="full",
+        queue=False,
+        concurrency_limit=None
+    )
+
+    generateBtn.click(
+        fn=generate_image,
+        inputs=[prompt, seed, width, height, randomize_seed, num_inference_steps],
+        outputs=[result, seed, latency],
+        show_progress="full",
+        api_name="RealtimeFlux",
+        queue=False
+    )
+
+    def update_ui(realtime_enabled):
+        return {
+            prompt: gr.update(interactive=True),
+            generateBtn: gr.update(visible=not realtime_enabled)
+        }
+
+    realtime.change(
+        fn=update_ui,
+        inputs=[realtime],
+        outputs=[prompt, generateBtn],
+        queue=False,
+        concurrency_limit=None
+    )
+
+    def realtime_generation(*args):
+        if args[0]:  # If realtime is enabled
+            return next(generate_image(*args[1:]))
+
+    prompt.submit(
+        fn=generate_image,
+        inputs=[prompt, seed, width, height, randomize_seed, num_inference_steps],
+        outputs=[result, seed, latency],
+        show_progress="full",
+        queue=False,
+        concurrency_limit=None
+    )
+
+    for component in [prompt, width, height, num_inference_steps]:
+        component.input(
+            fn=realtime_generation,
+            inputs=[realtime, prompt, seed, width, height, randomize_seed, num_inference_steps],
+            outputs=[result, seed, latency],
+            show_progress="hidden",
+            trigger_mode="always_last",
+            queue=False,
+            concurrency_limit=None
+        )
 
+# Launch the app
+demo.launch()