Add more instructions

app.py

@@ -65,7 +65,16 @@ def zero_shot_multimodal(image, text, options):
     return {label: probs[idx].item() for idx, label in enumerate(options)}
 
 with demo:
-    gr.Markdown("Zero-Shot Image Classification, Text Classification or even Multimodal using same FLAVA model.")
+    gr.Markdown(
+    """
+    # Zero-Shot image, text or multimodal classification using the same FLAVA model
+
+    Click on one the examples provided to load them into the UI and "Classify".
+
+    - For text and multimodal classification, provide your 1) prompt with the word you want to be filled in as `[MASK]`, and 2) possible options to be ranked separated by `;`.  
+    - For image classification, provide class options to be ranked separated by `;`.
+    """
+    )
     with gr.Tabs():
         with gr.TabItem("Zero-Shot Text Classification"):
             with gr.Row():