Update main.py

main.py

@@ -1,16 +1,10 @@
 
-
-
-
-
-
-
 from fastapi import FastAPI, File, UploadFile
 import os
 import shutil
 import tempfile
 from langchain_community.document_loaders import PyMuPDFLoader
-from LoadLLM import Loadllm
+from langchain_community.llms import CTransformers
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
 from langchain.chains import ConversationalRetrievalChain
@@ -43,7 +37,7 @@ async def PromptLLM(file: UploadFile = File(...)):
     db.save_local(DB_FAISS_PATH)
 
     # Load the language model
-    llm = Loadllm.load_llm()
+    llm = CTransformers(model="llama-2-7b-chat.ggmlv3.q4_1.bin", model_type="llama",config={'max_new_tokens': 1024,  'context_length': 2048, 'temperature': 0.01})
 
     # Create a conversational chain
     chain = ConversationalRetrievalChain.from_llm(llm=llm, retriever=db.as_retriever())