Update app.py

app.py

@@ -1,73 +1,240 @@
+import gradio as gr
+import numpy as np
+import random
 import spaces
 from diffusers import AuraFlowPipeline
 import torch
-import gradio as gr
+from gradio_imageslider import ImageSlider
 
-def initialize_auraflow_pipeline():
-    """Initialize and return the AuraFlowPipeline."""
-    pipeline = AuraFlowPipeline.from_pretrained(
-        "fal/AuraFlow-v0.3",
-        torch_dtype=torch.float16,
-        variant="fp16",
-    ).to("cuda")
-    return pipeline
+device = "cuda" if torch.cuda.is_available() else "cpu"
 
-@spaces.GPU(duration=95)
-def generate_image(pipeline, prompt, width, height, num_inference_steps, seed, guidance_scale):
-    """Generate an image using the AuraFlowPipeline."""
-    generator = torch.Generator().manual_seed(seed)
+#torch.set_float32_matmul_precision("high")
 
-    image = pipeline(
-        prompt=prompt,
-        width=width,
-        height=height,
-        num_inference_steps=num_inference_steps,
-        generator=generator,
-        guidance_scale=guidance_scale,
-    ).images[0]
+#torch._inductor.config.conv_1x1_as_mm = True
+#torch._inductor.config.coordinate_descent_tuning = True
+#torch._inductor.config.epilogue_fusion = False
+#torch._inductor.config.coordinate_descent_check_all_directions = True
 
-    return image
+#pipe_v1 = AuraFlowPipeline.from_pretrained(
+#	"fal/AuraFlow",
+#    torch_dtype=torch.float16
+#).to("cuda")
 
-# Initialize the pipeline once
-auraflow_pipeline = initialize_auraflow_pipeline()
+pipe_v2 = AuraFlowPipeline.from_pretrained(
+    "fal/AuraFlow-v0.2",
+    torch_dtype=torch.float16
+).to("cuda")
 
-# Gradio interface
-def gradio_generate_image(prompt, width, height, num_inference_steps, seed, guidance_scale):
-    return generate_image(auraflow_pipeline, prompt, width, height, num_inference_steps, seed, guidance_scale)
+pipe = AuraFlowPipeline.from_pretrained(
+	"fal/AuraFlow-v0.3",
+    torch_dtype=torch.float16
+).to("cuda")
+#pipe.transformer.to(memory_format=torch.channels_last)
+#pipe.transformer = torch.compile(pipe.transformer, mode="reduce-overhead", fullgraph=True)
+#pipe.transformer.to(memory_format=torch.channels_last)
+#pipe.vae.to(memory_format=torch.channels_last)
 
-# Create Gradio Blocks
-with gr.Blocks(theme="bethecloud/storj_theme") as demo:
-    gr.HTML(
-        """
-    <h1 style='text-align: center'>
-    AuraFlow v0.3
-    </h1>
-    """)
-    gr.HTML(
-        """
-        <h3 style='text-align: center'>
-        Follow me for more!
-        <a href='https://twitter.com/kadirnar_ai' target='_blank'>Twitter</a> | <a href='https://github.com/kadirnar' target='_blank'>Github</a> | <a href='https://www.linkedin.com/in/kadir-nar/' target='_blank'>Linkedin</a>  | <a href='https://www.huggingface.co/kadirnar/' target='_blank'>HuggingFace</a>
-        </h3>
-        """)
-    with gr.Row():
-        with gr.Column():
-            prompt_input = gr.Textbox(label="Prompt", placeholder="Enter your image prompt here...")
-            width_input = gr.Slider(minimum=256, maximum=1536, step=64, value=1024, label="Width")
-            height_input = gr.Slider(minimum=256, maximum=1536, step=64, value=1024, label="Height")
-            steps_input = gr.Slider(minimum=10, maximum=100, step=1, value=20, label="Number of Inference Steps")
-            seed_input = gr.Number(label="Seed", value=1)
-            guidance_input = gr.Slider(minimum=1, maximum=10, step=0.1, value=3.5, label="Guidance Scale")
-            generate_btn = gr.Button("Generate Image")
+#pipe.transformer = torch.compile(pipe.transformer, mode="max-autotune", fullgraph=True)
+#pipe.vae.decode = torch.compile(pipe.vae.decode, mode="max-autotune", fullgraph=True)
+
+MAX_SEED = np.iinfo(np.int32).max
+MAX_IMAGE_SIZE = 1024
+
+@spaces.GPU()
+def infer_example(prompt, negative_prompt="", seed=42, randomize_seed=False, width=1024, height=1024, guidance_scale=5.0, num_inference_steps=28, model_version="0.2", comparison_mode=False, progress=gr.Progress(track_tqdm=True)):
+    generator = torch.Generator().manual_seed(seed)
+    image = pipe(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width = width,
+            height = height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+    ).images[0] 
+    return image, seed
+    
+@spaces.GPU(duration=95)
+def infer(prompt,
+          negative_prompt="",
+          seed=42,
+          randomize_seed=False,
+          width=1024,
+          height=1024,
+          guidance_scale=5.0,
+          num_inference_steps=28,
+          model_version="0.3",
+          comparison_mode=False,
+          progress=gr.Progress(track_tqdm=True)
+         ):
+    
+    if randomize_seed:
+        seed = random.randint(0, MAX_SEED)
         
-        with gr.Column():
-            image_output = gr.Image(label="Generated Image")
+    generator = torch.Generator().manual_seed(seed)
+    if(comparison_mode):
+        image_1 = pipe_v2(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width=width,
+            height=height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+        ).images[0] 
+        generator = torch.Generator().manual_seed(seed)
+        image_2 = pipe(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width=width,
+            height=height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+        ).images[0]
+        return gr.update(visible=False), gr.update(visible=True, value=(image_1, image_2)), seed
+    if(model_version == "0.1"):
+        image = pipe_v1(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width=width,
+            height=height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+        ).images[0]
+    elif(model_version == "0.2"):
+        image = pipe_v2(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width=width,
+            height=height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+        ).images[0]
+    else:
+        image = pipe(
+            prompt = prompt, 
+            negative_prompt = negative_prompt,
+            width=width,
+            height=height,
+            guidance_scale = guidance_scale, 
+            num_inference_steps = num_inference_steps, 
+            generator = generator
+        ).images[0]
+        
+    return gr.update(visible=True, value=image), gr.update(visible=False), seed
+
+examples = [
+    "A photo of a lavender cat",
+    "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k",
+    "An astronaut riding a green horse",
+    "A delicious ceviche cheesecake slice",
+]
+
+css="""
+#col-container {
+    margin: 0 auto;
+    max-width: 520px;
+}
+"""
+
+with gr.Blocks(css=css) as demo:
     
-    generate_btn.click(
-        fn=gradio_generate_image,
-        inputs=[prompt_input, width_input, height_input, steps_input, seed_input, guidance_input],
-        outputs=image_output
+    with gr.Column(elem_id="col-container"):
+        gr.Markdown(f"""
+        # AuraFlow 0.3
+        Demo of the [AuraFlow 0.3](https://huggingface.co/fal/AuraFlow-v0.3) 6.8B parameters open source diffusion transformer model
+        [[blog](https://blog.fal.ai/auraflow/)] [[model](https://huggingface.co/fal/AuraFlow)] [[fal](https://fal.ai/models/fal-ai/aura-flow)]
+        """)
+        
+        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)
+        result_compare = ImageSlider(visible=False, label="Left 0.2, Right 0.3")
+        comparison_mode = gr.Checkbox(label="Comparison mode", info="Compare v0.2 with v0.3", value=False)
+        with gr.Accordion("Advanced Settings", open=False):
+            
+            model_version = gr.Dropdown(
+                ["0.2", "0.3"], label="Model version", value="0.3"
+            )
+            
+            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=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=0.0,
+                    maximum=10.0,
+                    step=0.1,
+                    value=5.0,
+                )
+                
+                num_inference_steps = gr.Slider(
+                    label="Number of inference steps",
+                    minimum=1,
+                    maximum=50,
+                    step=1,
+                    value=28,
+                )
+        
+        gr.Examples(
+            examples = examples,
+            fn = infer_example,
+            inputs = [prompt],
+            outputs = [result, seed],
+            cache_examples="lazy"
+        )
+
+    gr.on(
+        triggers=[run_button.click, prompt.submit, negative_prompt.submit],
+        fn = infer,
+        inputs = [prompt, negative_prompt, seed, randomize_seed, width, height, guidance_scale, num_inference_steps, model_version, comparison_mode],
+        outputs = [result, result_compare, seed]
     )
 
-# Launch the Gradio interface
-demo.launch()
+demo.queue().launch()