add batch evaluation

pages/batch_evaluation.py

@@ -2,6 +2,8 @@ from dotenv import load_dotenv
 import gradio as gr
 import random
 
+import markdown
+
 from utils.model import Model
 from utils.data import dataset
 from utils.metric import metric_rouge_score
@@ -10,6 +12,37 @@ from pages.summarization_playground import model, generate_answer
 
 load_dotenv()
 
+def display_results(response_list):
+    html_output = ""
+    
+    for i, item in enumerate(response_list, 1):
+        dialogue = item['dialogue']
+        summary = item['summary']
+        response = item['response']
+        rouge_score = item['metric_score']['rouge_score']
+        
+        html_output += f"""
+        <details>
+        <summary>Response {i} (Rouge Score: {rouge_score:.2f})</summary>
+        <div style="display: flex; justify-content: space-between;">
+            <div style="width: 30%;">
+                <h3>Dialogue</h3>
+                {markdown.markdown(dialogue)}
+            </div>
+            <div style="width: 30%;">
+                <h3>Summary</h3>
+                {markdown.markdown(summary)}
+            </div>
+            <div style="width: 30%;">
+                <h3>Response</h3>
+                {markdown.markdown(response)}
+            </div>
+        </div>
+        </details>
+        """
+    
+    return html_output
+
 def process(seed, model_selection, prompt, num=10):
     random.seed(seed)
     response_list = []
@@ -32,11 +65,12 @@ def process(seed, model_selection, prompt, num=10):
             }
         )
 
-    return response_list
+    return display_results(response_list)
+
 
 def create_batch_evaluation_interface():
     with gr.Blocks() as demo:
-        gr.Markdown("## Here are evaluation setups")
+        gr.Markdown("## Here are evaluation setups. It will randomly sample 10 data points to generate and evaluate. Show results once finished.")
         
         with gr.Row():
             seed = gr.Number(value=8, info="pick your favoriate random seed", precision=0)

pages/leaderboard.py

@@ -41,7 +41,7 @@ def create_leaderboard():
         
         sort_by = gr.Dropdown(list(df.columns), label="Sort by", value="Rank")
         
-        stats = gr.Markdown("**Performance**\n\n**methods**: 4,   **questions**: 150")
+        gr.Markdown("**Performance**\n\n**methods**: 4,   **questions**: 150")
         
         leaderboard = gr.HTML(update_leaderboard("Rank"), elem_id="leaderboard")