Update summarize.py

summarize.py

@@ -63,8 +63,8 @@ def summarize_and_score(ids, mask, model, tokenizer, **kwargs):
         remove_invalid_values=True,
     )
     score = round(summary_pred_ids.sequences_scores.cpu().numpy()[0], 4)
-
-    return summary, score
+    len_res = len(summary_pred_ids.sequences.cpu().numpy()[0])
+    return summary, score, len_res
 
 
 def summarize_via_tokenbatches(
@@ -111,7 +111,7 @@ def summarize_via_tokenbatches(
 
     for _id, _mask in zip(in_id_arr, att_arr):
 
-        result, score = summarize_and_score(
+        result, score, l = summarize_and_score(
             ids=_id,
             mask=_mask,
             model=model,
@@ -119,10 +119,12 @@ def summarize_via_tokenbatches(
             **kwargs,
         )
         score = round(float(score), 4)
+        rate = round(float(len(_id)-l)/len(_id)),3)
         _sum = {
             "input_tokens": _id,
             "summary": result,
             "summary_score": score,
+            "compression_rate": rate,
         }
         gen_summaries.append(_sum)
         print(f"\t{result[0]}\nScore:\t{score}")