update

pages/2_batch_evaluation.py

@@ -3,19 +3,6 @@ import streamlit as st
 from util.evaluator import evaluator, write_evaluation_commentary
 import os
 
-# Predefined examples
-examples = {
-    'good': {
-        'question': "What causes rainbows to appear in the sky?",
-        'explanation': "Rainbows appear when sunlight is refracted, dispersed, and reflected inside water droplets in the atmosphere, resulting in a spectrum of light appearing in the sky."
-    },
-    'bad': {
-        'question': "What causes rainbows to appear in the sky?",
-        'explanation': "Rainbows happen because light in the sky gets mixed up and sometimes shows colors when it's raining or when there is water around."
-    }
-}
-
-# Function to check password
 def check_password():
     with st.sidebar:
         password_input = st.text_input("Enter Password:", type="password")
@@ -27,83 +14,45 @@ def check_password():
             else:
                 st.error("Incorrect Password, please try again.")
 
-# Function to evaluate batch data
-def evaluate_batch(uploaded_file):
+def batch_evaluate(uploaded_file):
+    # Read the uploaded CSV file into DataFrame
     df = pd.read_csv(uploaded_file)
-    eval_instance = evaluator(model_name=st.session_state['model_name'])
+    eval_instance = evaluator('gpt4-1106')  # Using fixed model name for simplicity
     results = []
 
+    # Process each row in the DataFrame
     for _, row in df.iterrows():
-        scores = eval_instance(row['question'], row['explanation'])
-        commentary = write_evaluation_commentary(scores)
-        result = {
-            'Question': row['question'],
-            'Explanation': row['explanation'],
-            **{c['Principle']: c['Score'] for c in commentary}
-        }
-        results.append(result)
+        question = row['question']
+        explanation = row['explanation']
+        scores = eval_instance(question, explanation)  # Evaluate using the evaluator
+        commentary_details = write_evaluation_commentary(scores)  # Generate commentary based on scores
+        results.append({
+            'Question': question,
+            'Explanation': explanation,
+            **{detail['Principle']: detail['Score'] for detail in commentary_details}
+        })
 
     return pd.DataFrame(results)
 
-# Main app logic
-def main():
-    st.title('Natural Language Explanation Demo')
-
-    model_name = st.selectbox('Select a model:', ['gpt4-1106', 'gpt35-1106'])
-    st.session_state['model_name'] = model_name  # Save model name to session state for use in batch processing
+st.title('Natural Language Explanation Demo')
 
-    input_type = st.radio("Choose input type:", ('Use predefined example', 'Enter your own', 'Upload CSV for batch evaluation'))
+if 'password_verified' not in st.session_state or not st.session_state['password_verified']:
+    check_password()
+else:
+    st.sidebar.success("Password Verified. Proceed with the demo.")
+    uploaded_file = st.file_uploader("Upload CSV file with 'question' and 'explanation' columns", type=['csv'])
 
-    if input_type == 'Use predefined example':
-        example_type = st.radio("Select an example type:", ('good', 'bad'))
-        question = examples[example_type]['question']
-        explanation = examples[example_type]['explanation']
-    elif input_type == 'Enter your own':
-        question = st.text_input('Enter your question:', '')
-        explanation = st.text_input('Enter your explanation:', '')
-    else:
-        uploaded_file = st.file_uploader("Upload a CSV file", type='csv')
-        if uploaded_file and st.button('Evaluate Batch'):
-            result_df = evaluate_batch(uploaded_file)
+    if uploaded_file is not None:
+        if st.button('Evaluate Explanations'):
+            result_df = batch_evaluate(uploaded_file)
             st.write('### Evaluated Results')
             st.dataframe(result_df)
-            csv = result_df.to_csv(index=False)
-            st.download_button(
-                label="Download evaluated results as CSV",
-                data=csv,
-                file_name='batch_evaluation_results.csv',
-                mime='text/csv'
-            )
-            return
-
-    if st.button('Evaluate Explanation'):
-        if question and explanation:
-            eval_instance = evaluator(model_name)
-            scores = eval_instance(question, explanation)
-            st.write('### Scores')
-            details = write_evaluation_commentary(scores)
-            df = pd.DataFrame(details)
-            st.write(df)
-            data = {
-                'Question': question,
-                'Explanation': explanation,
-                **{detail['Principle']: detail['Score'] for detail in details}
-            }
-            df = pd.DataFrame([data])
 
-            # Convert DataFrame to CSV for download
-            csv = df.to_csv(index=False)
+            # Create a CSV download link
+            csv = result_df.to_csv(index=False)
             st.download_button(
-                label="Download evaluation as CSV",
+                label="Download evaluation results as CSV",
                 data=csv,
-                file_name='evaluation.csv',
+                file_name='evaluated_results.csv',
                 mime='text/csv',
             )
-        else:
-            st.error('Please enter both a question and an explanation to evaluate.')
-
-if __name__ == '__main__':
-    if 'password_verified' not in st.session_state or not st.session_state['password_verified']:
-        check_password()
-    else:
-        main()