added

app.py

@@ -206,6 +206,8 @@ def load_queries(dataset_name):
 
 
 def evaluate(qrels, results, k_values):
+    qrels = {str(k): {str(k2): v2 for k2, v2 in v.items()} for k, v in qrels.items()}
+    results = {str(k): {str(k2): v2 for k2, v2 in v.items()} for k, v in results.items()}
     evaluator = pytrec_eval.RelevanceEvaluator(
         qrels, {f"ndcg_cut.{k}" for k in k_values} | {f"recall.{k}" for k in k_values}
     )
@@ -232,13 +234,16 @@ def run_evaluation(dataset, postfix):
     logger.info(f"Sample input text: {input_texts[0]}")
 
     q_reps = model.encode(input_texts)
+    logger.info(f"Encoded query first five: {q_reps[0][:5]}")
     logger.info(f"Encoded query representations shape: {q_reps.shape}")
 
     all_scores, psg_indices = search_queries(dataset, q_reps)
     
-    results = {qid: dict(zip(doc_ids, map(float, scores))) 
-               for qid, scores, doc_ids in zip(q_lookups[dataset].keys(), all_scores, psg_indices)}
-    
+    results = {str(qid): {str(doc_id): float(score) for doc_id, score in zip(doc_ids, scores)}
+           for qid, scores, doc_ids in zip(q_lookups[dataset].keys(), all_scores, psg_indices)}
+    qrels[dataset] = {str(qid): {str(doc_id): rel for doc_id, rel in rels.items()}
+                  for qid, rels in qrels[dataset].items()}
+
     logger.info(f"Number of results: {len(results)}")
     logger.info(f"Sample result: {list(results.items())[0]}")