Update app.py

app.py

@@ -67,7 +67,7 @@ class Retriever:
   def load_chunks(self):
     self.text = self.extract_text_from_pdf(self.file_path)
     text_splitter = RecursiveCharacterTextSplitter(
-        chunk_size=300,
+        chunk_size=150,
         chunk_overlap=20,
         length_function=self.token_len,
         separators=["Section", "\n\n", "\n", ".", " ", ""]
@@ -86,7 +86,7 @@ class Retriever:
     self.index.add(self.token_embeddings)
 
   def retrieve_top_k(self, query_prompt, k=10):
-    encoded_query = self.question_tokenizer(query_prompt, return_tensors="pt", max_length=300, truncation=True, padding=True).to(device)
+    encoded_query = self.question_tokenizer(query_prompt, return_tensors="pt", truncation=True, padding=True).to(device)
 
     with torch.no_grad():
         model_output = self.question_model(**encoded_query)
@@ -99,6 +99,7 @@ class Retriever:
 
     return retrieved_texts
 
+
 class RAG:
     def __init__(self,
                  file_path,
@@ -134,7 +135,7 @@ class RAG:
       return answer
 
     def extractive_query(self, question):
-      context = self.retriever.retrieve_top_k(question, k=4)
+      context = self.retriever.retrieve_top_k(question, k=7)
       
       inputs = self.generator_tokenizer(question, ". ".join(context), return_tensors="pt", truncation=True, max_length=150, padding="max_length")
       with torch.no_grad():