Potential Fix - title bug

summarizer.py

@@ -91,8 +91,9 @@ class BARTSummarizer:
         # Summarize each cluster
         summaries_with_title = []
         for cluster in clustered_sentences:
+            title = self.title_model.summarize(cluster)
             summary = self.chunk_summarize(cluster, auto=True)
-            summary_with_title = "#### " + self.title_model.summarize(summary) + "\n" + summary
+            summary_with_title = "#### " + title + "\n" + summary
             summaries_with_title.append(summary_with_title)
 
         # Combine the summaries to get the final summary for the entire input

title_generator.py

@@ -9,6 +9,6 @@ class T5Summarizer:
         inputs = ["summarize: " + text]
         max_input_length = self.tokenizer.model_max_length
         inputs = self.tokenizer(inputs, max_length=max_input_length, truncation=True, return_tensors="tf")
-        output = self.model.generate(**inputs, num_beams=8, do_sample=True, min_length=1, max_length=10, early_stopping=True)
+        output = self.model.generate(**inputs, num_beams=12, do_sample=True, min_length=2, max_length=8)
         summary = self.tokenizer.batch_decode(output, skip_special_tokens=True)[0]
         return summary