Update README.md

README.md

@@ -96,6 +96,7 @@ import os
 from PIL import Image
 
 # Path to your dataset CSV and image folder
+# get our sample from here https://huggingface.co/vector-institute/Llama3.2-Multimodal-Newsmedia-Bias-Detector/tree/main/sampled-data
 dataset_csv_path = 'sample_dataset.csv'
 image_folder_path = 'sampled_images'
 
@@ -224,7 +225,7 @@ def generate_prediction(sample):
 samples = prepare_samples_from_dataframe(df, image_folder_path)
 
 # Limit to a subset if desired
-samples_to_infer = samples[:5]  # For example, take the first 5 samples
+samples_to_infer = samples[:900]  # For example, take the first 5 samples
 
 # Run inference and collect results
 results = []
@@ -232,26 +233,29 @@ results = []
 for sample in samples_to_infer:
     assistant_reply = generate_prediction(sample)
     predicted_label = extract_classification(assistant_reply)
-
+        # Get the first_paragraph from the original DataFrame
+    original_row = df[df['unique_id'] == sample['unique_id']].iloc[0]
+    first_paragraph = original_row['first_paragraph']
     # Collect results
     result = {
-        'unique_id': sample['unique_id'],
-        'assistant_reply': assistant_reply,
-        'predicted_label': predicted_label,
-        # Add any other fields you need from the sample or DataFrame
+    'unique_id': sample['unique_id'],
+    'first_paragraph': first_paragraph,  # Add this line
+    'assistant_reply': assistant_reply,
+    'predicted_label': predicted_label,
     }
     results.append(result)
 
     # Display the results
     print(f"Sample ID: {sample['unique_id']}")
+    print(f"text: {first_paragraph}")
     print("Assistant's Reply:")
     print(assistant_reply)
-    print(f"Predicted Label: {predicted_label}")
+    #print(f"Predicted Label: {predicted_label}")
     print("-" * 50)
 
 results_df = pd.DataFrame(results)
 # Save to CSV if desired
-results_df.to_csv('inference_results.csv', index=False)
+results_df.to_csv('llama-vision-ift-inference_results.csv', index=False)
 ```
 ### Example Output