changed to small in tokenization and num_beams=2 now

chatbot.py

@@ -21,7 +21,7 @@ def load_models():
     print("Loading models...")
     
     retriever = SentenceTransformer("flax-sentence-embeddings/all_datasets_v3_mpnet-base")
-    tokenizer = T5Tokenizer.from_pretrained('t5-base')
+    tokenizer = T5Tokenizer.from_pretrained('t5-small')
     generator = T5ForConditionalGeneration.from_pretrained('t5-base').to(device)
 
     return retriever, generator, tokenizer
@@ -57,7 +57,7 @@ def predict(query: QueryInput):
         return {"response": "The topic is not covered in the student manual."}
 
     inputs = tokenizer.encode(formatted_query, return_tensors="pt", max_length=512, truncation=True).to(device)
-    ids = generator.generate(inputs, num_beams=4, min_length=10, max_length=60, repetition_penalty=1.2)
+    ids = generator.generate(inputs, num_beams=2, min_length=10, max_length=60, repetition_penalty=1.2)
     answer = tokenizer.decode(ids[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
 
     return {"response": answer}