Update description

app.py

@@ -114,6 +114,13 @@ def predict(question, kg_model, qa_model, max_input_len, max_output_len, m, top_
     output += f'Knowledge selected to make the prediction: {result["selected_knowledge"]}\n'
     return output
 
+description = '''This is a demo for the paper, <a href="https://arxiv.org/pdf/2210.03078.pdf" target="_blank">Rainier: Reinforced Knowledge Introspector for Commonsense Question Answering</a>, presented in EMNLP 2022.
+[<a href="https://github.com/liujch1998/rainier" target="_blank">Code</a>] [<a href="https://huggingface.co/liujch1998/rainier-large" target="_blank">Model</a>]
+This demo is made & maintained by <a href="https://liujch1998.github.io/" target="_blank">Jiacheng (Gary) Liu</a>.
+
+Rainier is a knowledge-generating model that enhances the commonsense QA capability of a QA model.
+To try this model, select an example question, or write your own question in the suggested format.'''
+
 examples = [
     'If the mass of an object gets bigger what will happen to the amount of matter contained within it? \\n (A) gets bigger (B) gets smaller',
     'What would vinyl be an odd thing to replace? \\n (A) pants (B) record albums (C) record store (D) cheese (E) wallpaper',
@@ -123,22 +130,23 @@ examples = [
     'Causes bad breath and frightens blood-suckers \\n (A) tuna (B) iron (C) trash (D) garlic (E) pubs',
 ]
 
-input_question = gr.Dropdown(
-    choices=examples,
-    label='Question:',
+input_question = gr.Dropdown(choices=examples, label='Question:',
     info='A multiple-choice commonsense question. Please follow the UnifiedQA input format: "{question} \\n (A) ... (B) ... (C) ..."',
 )
 input_kg_model = gr.Textbox(label='Knowledge generation model:', value='liujch1998/rainier-large', interactive=False)
 input_qa_model = gr.Textbox(label='QA model:', value='allenai/unifiedqa-t5-large', interactive=False)
-input_max_input_len = gr.Number(label='Max question length:', value=256, precision=0)
-input_max_output_len = gr.Number(label='Max knowledge length:', value=32, precision=0)
-input_m = gr.Slider(label='Number of generated knowledges:', value=10, mininum=1, maximum=20, step=1)
-input_top_p = gr.Slider(label='Top_p for knowledge generation:', value=0.5, mininum=0.0, maximum=1.0, step=0.05)
+input_max_input_len = gr.Number(label='Max number of tokens in question:', value=256, precision=0)
+input_max_output_len = gr.Number(label='Max number of tokens in knowledge:', value=32, precision=0)
+input_m = gr.Slider(label='Number of generated knowledges:', value=10, mininum=1, maximum=20, step=1,
+    info='The actual number of generated knowledges may be less than this number due to possible duplicates.',
+)
+input_top_p = gr.Slider(label='top_p for knowledge generation:', value=0.5, mininum=0.0, maximum=1.0, step=0.05)
 output_text = gr.Textbox(label='Output', interactive=False)
 
 gr.Interface(
     fn=predict,
     inputs=[input_question, input_kg_model, input_qa_model, input_max_input_len, input_max_output_len, input_m, input_top_p],
     outputs=output_text,
-    title="Rainier",
+    title="Rainier Demo",
+    description=description,
 ).launch()