Update src/models/summarization.py

src/models/summarization.py

@@ -1,8 +1,6 @@
-
-from transformers import BartTokenizer
+from transformers import BartTokenizer, BartForConditionalGeneration
 import torch
 import streamlit as st
-import pickle
 
 class Summarizer:
     def __init__(self):
@@ -12,18 +10,26 @@ class Summarizer:
     def load_model(self):
         try:
             self.tokenizer = BartTokenizer.from_pretrained('facebook/bart-base')
-            with open('bart_ami_finetuned.pkl', 'rb') as f:
-                self.model = pickle.load(f)
+            self.model = torch.load('bart_ami_finetuned.pkl')
+            self.model.to(torch.device('cuda' if torch.cuda.is_available() else 'cpu'))
             return self.model
         except Exception as e:
             st.error(f"Error loading summarization model: {str(e)}")
             return None
 
-    def process(self, text: str, max_length: int = 130, min_length: int = 30):
+    def process(self, text: str, max_length: int = 150, min_length: int = 40):
         try:
             inputs = self.tokenizer(text, return_tensors="pt", max_length=1024, truncation=True)
-            summary_ids = self.model.generate(inputs["input_ids"], max_length=max_length, min_length=min_length)
-            return self.tokenizer.decode(summary_ids[0], skip_special_tokens=True)
+            inputs = {key: value.to(self.model.device) for key, value in inputs.items()}
+            summary_ids = self.model.generate(
+                inputs["input_ids"],
+                max_length=max_length,
+                min_length=min_length,
+                num_beams=4,
+                length_penalty=2.0
+            )
+            summary = self.tokenizer.decode(summary_ids[0], skip_special_tokens=True)
+            return summary
         except Exception as e:
             st.error(f"Error in summarization: {str(e)}")
-            return None
+            return None