Update app.py

app.py

@@ -23,7 +23,7 @@ def diffusion_schedule(diffusion_times, min_signal_rate, max_signal_rate):
 
 def generate_images(diffusion_steps, stochasticity, min_signal_rate, max_signal_rate):
     step_size = 1.0 / diffusion_steps
-    initial_noise = tf.random.normal(shape=(num_images, image_size, image_size, 3))
+    initial_noise = tf.random.normal(shape=(num_images, image_size, image_size, 3), seed=42)
     
     # reverse diffusion
     noisy_images = initial_noise
@@ -57,13 +57,21 @@ def generate_images(diffusion_steps, stochasticity, min_signal_rate, max_signal_
     generated_images = tf.reshape(generated_images, (num_rows * plot_image_size, num_cols * plot_image_size, 3))
     return generated_images.numpy()
 
+inputs = [
+    gr.inputs.Slider(1, 20, step=1, default=10, label="Diffusion steps"),
+    gr.inputs.Slider(0.0, 1.0, step=0.05, default=0.0, label="Stochasticity (η in the paper)"),
+    gr.inputs.Slider(0.02, 0.10, step=0.01, default=0.02, label="Minimal signal rate"),
+    gr.inputs.Slider(0.80, 0.95, step=0.01, default=0.95, label="Maximal signal rate"),
+]
+output = gr.outputs.Image(label="Generated images")
+examples = [[2, 0.0, 0.02, 0.95], [10, 0.0, 0.02, 0.95], [20, 1.0, 0.02, 0.95]]
+title = "Denoising Diffusion Implicit Models"
+article = "<div style='text-align: center;'>Keras code example and demo by <a href='https://www.linkedin.com/in/andras-beres-789190210' target='_blank'>András Béres</a></div>"
 gr.Interface(
     generate_images,
-    inputs=[
-        gr.inputs.Slider(1, 20, step=1, default=10, label="Diffusion steps"),
-        gr.inputs.Slider(0.0, 1.0, step=0.05, default=0.0, label="Stochasticity"),
-        gr.inputs.Slider(0.02, 0.10, step=0.01, default=0.02, label="Minimal signal rate"),
-        gr.inputs.Slider(0.80, 0.95, step=0.01, default=0.95, label="Maximal signal rate"),
-    ],
-    outputs="image",
+    inputs=inputs,
+    outputs=output,
+    examples=examples,
+    title=title,
+    article=article,
 ).launch()