Updated with new example

app.py

@@ -29,7 +29,7 @@ try:
     # Load pre-computed vocabulary scores and names
     vocab_scores = np.load(VOCAB_SCORES_PATH)
     with open(VOCAB_NAMES_PATH, 'r') as f:
-        vocab_names = [line.strip() for line in f.readlines()]
+        vocab_names = [line.strip().lower() for line in f.readlines()]
 
 except FileNotFoundError as e:
     print(f"ERROR: A required file was not found: {e.filename}")
@@ -99,6 +99,7 @@ def predict(input_img, top_k, concept, neg_concept, max_strength):
     # --- Part B: Concept Manipulation ---
 
     # Validate the user-provided concept
+    concept = concept.lower().strip()
     if concept not in vocab_names:
         raise gr.Error(f"Concept '{concept}' not found in vocabulary. Please choose a valid concept.")
     
@@ -113,7 +114,7 @@ def predict(input_img, top_k, concept, neg_concept, max_strength):
     if not neg_concept:
         neg_concept_prompt = f"a photo without {concept}"
     else:
-        neg_concept_prompt = f"a photo with {neg_concept}"
+        neg_concept_prompt = f"a photo with {neg_concept.lower().strip()}"
     
     pos_concept_prompt = f"a photo with {concept}"
     
@@ -187,6 +188,19 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Matryoshka Sparse Autoencoder (MSA
         "Upload an image to see its top activating concepts from a sparse autoencoder. Then, choose a concept (from `clip_disect_20k.txt`) to visualize how manipulating its corresponding concept magnitude affects the image representation."
     )
     
+    gr.Examples(
+        examples=[
+            ["./bird.jpg", 10, "birds", "", 10.0],
+            ["./statue.jpg", 10, "statue", "humans", 10.0],
+        ],
+        # NOTE: You might need to create placeholder images 'bird.jpg' and 'statue.jpg'
+        # in your directory for the examples to load correctly.
+        inputs=[image_input, top_k_slider, concept_input, neg_concept_input, max_strength_slider],
+        outputs=[output_image, output_bar_plot, output_line_plot],
+        fn=predict,
+        cache_examples=True # Set to True for faster loading on HF Spaces
+    )
+    
     with gr.Row():
         with gr.Column(scale=1):
             # Input controls
@@ -205,16 +219,6 @@ with gr.Blocks(theme=gr.themes.Soft(), title="Matryoshka Sparse Autoencoder (MSA
             output_bar_plot = gr.Plot(label="Top Activating Concepts")
             output_line_plot = gr.Plot(label="Concept Manipulation Analysis")
 
-    gr.Examples(
-        examples=[
-            ["bird.jpg", 10, "birds", "", 10.0],
-        ],
-        inputs=[image_input, top_k_slider, concept_input, neg_concept_input, max_strength_slider],
-        outputs=[output_image, output_bar_plot, output_line_plot],
-        fn=predict,
-        cache_examples=True # Set to True for faster loading on HF Spaces
-    )
-
     # Wire up the button to the function
     submit_btn.click(
         fn=predict,