reduce processing time

app.py

@@ -44,16 +44,20 @@ async def summarize(req: Request, body: SummarizeRequest):
     attention_mask = encoding["attention_mask"].to(device)
 
     # Sinh tóm tắt với cấu hình ổn định
+    # outputs = model.generate(
+    #     input_ids=input_ids,
+    #     attention_mask=attention_mask,
+    #     max_length=128,
+    #     num_beams=1,
+    #     early_stopping=True,
+    #     no_repeat_ngram_size=2,
+    #     num_return_sequences=1
+    # )
     outputs = model.generate(
-        input_ids=input_ids,
-        attention_mask=attention_mask,
-        max_length=128,
-        num_beams=1,
-        early_stopping=True,
-        no_repeat_ngram_size=2,
-        num_return_sequences=1
+        input_ids=input_ids, attention_mask=attention_masks,
+        max_length=256,
+        early_stopping=True
     )
-
     summary = tokenizer.decode(outputs[0], skip_special_tokens=True, clean_up_tokenization_spaces=True)
 
     end_time = time.time()