use CLS token, tuning default params

app.py

@@ -17,7 +17,7 @@ def process_image(input_image, learning_rate, iterations):
     
     def get_encoder_activations(x):
         encoder_output = model.vit(x)
-        final_activations = encoder_output.last_hidden_state
+        final_activations = encoder_output.last_hidden_state[:,0,:]
         return final_activations
 
     image = input_image.convert('RGB')
@@ -47,8 +47,8 @@ iface = gr.Interface(
     fn=process_image,
     inputs=[
         gr.Image(type="pil"), 
-        gr.Number(value=0.01, label="Learning Rate"), 
-        gr.Number(value=1, label="Iterations")
+        gr.Number(value=4.0, label="Learning Rate"), 
+        gr.Number(value=4, label="Iterations")
     ],
     outputs=[gr.Image(type="numpy", label="ViT-Dreamed Image")]
 )