I have the following code:

from langchain_community.document_loaders.csv_loader import CSVLoader
from langchain.text_splitter import RecursiveCharacterTextSplitter
# from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_community.llms import CTransformers
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
import sys

# CTransformers for GGML models
# This script works relatively well with the small example
# but we have not tried it with the big one


DB_FAISS_PATH = "vectorstore/db_faiss"

# # datafile = "../expdata/output_2024-09-06_17-44-59.txt"
# datafile = "../expdata/example.txt"
datafile = "../expdata/output_2024-09-06_17-44-59.csv"
# datafile = "../expdata/example.csv"

loader = CSVLoader(file_path=datafile, encoding="utf-8", csv_args={'delimiter': ','})
data = loader.load()

# Split the text into Chunks
text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=20)  #Put 100 later
text_chunks = text_splitter.split_documents(data)

print(f" Text chunks len {len(text_chunks)}")
print("----------------")
print(text_chunks[0])
print("-----------------")
# Download Sentence Transformers Embedding From Hugging Face
embeddings = HuggingFaceEmbeddings(model_name = 'sentence-transformers/all-MiniLM-L6-v2')

# what we called vectordb in the other script
docsearch = FAISS.from_documents(text_chunks, embeddings)
docsearch.save_local(DB_FAISS_PATH)

MAX_TOKENS= 1024

llm = CTransformers(model="../models/llama-2-7b-chat.ggmlv3.q4_0.bin",
                    model_type="llama",
                    max_new_tokens=MAX_TOKENS,
                    temperature=0.1)

qa = ConversationalRetrievalChain.from_llm(llm, retriever=docsearch.as_retriever())


while True:
    chat_history = []
    #query = "What is the value of  GDP per capita of Finland provided in the data?"
    query = input(f"Input Prompt: ")
    if query == 'exit':
        print('Exiting')
        sys.exit()
    if query == '':
        continue
    result = qa({"question":query, "chat_history":chat_history})
    print(f"History {chat_history}")
    print("RESPONSE: ", result['answer'])

When I run this code with a small csv file (10 rows, one sentence) (example.csv) the script works reasonably well. (and this was when MAX_TOKENS was 512)

However when I run the script with a bigger csv file (output_2024-09-06_17-44-59.csv), I got the following:

Number of tokens (525) exceeded maximum context length (512)

This was when MAX_TOKENS was 512 but also now that it is 1024 ( and there is no 512 value anywhere)

Why is this happening and how can I solve this?