Initial Draft

app.py

@@ -1,8 +1,6 @@
 import streamlit as st
 import os
 import requests
-import time
-import sys
 
 from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import CharacterTextSplitter
@@ -10,7 +8,7 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain.vectorstores.faiss import FAISS
 
 from langchain.prompts.prompt import PromptTemplate
-from langchain_community.llms import CTransformers
+from langchain_community.llms import LlamaCpp
 from langchain.chains import RetrievalQA
 
 # Upload pdf file into 'pdf-data' folder if it does not exist
@@ -40,10 +38,10 @@ def fn_create_vector_db(mv_pdf_input_file, mv_processing_message):
 
     lv_file_name = mv_pdf_input_file.name[:-4] + ".vectorstore"
 
-    if not os.path.exists("vectordb/fiaas"):
-        os.makedirs("vectordb/fiaas")
+    if not os.path.exists(os.path.join("vectordb","fiaas")):
+        os.makedirs(os.path.join("vectordb","fiaas"))
     
-    lv_temp_file_path = os.path.join("vectordb/fiaas",lv_file_name)
+    lv_temp_file_path = os.path.join(os.path.join("vectordb","fiaas"),lv_file_name)
     lv_embeddings = HuggingFaceEmbeddings(
                                             model_name="sentence-transformers/all-mpnet-base-v2",
                                             model_kwargs={'device': 'cpu'}
@@ -111,6 +109,8 @@ def fn_download_llm_models(mv_selected_model, mv_processing_message):
 
     if mv_selected_model == 'microsoft/phi-2':
         lv_download_url = "https://huggingface.co/TheBloke/phi-2-GGUF/resolve/main/phi-2.Q2_K.gguf"
+    elif mv_selected_model == 'google/gemma-2b':
+        lv_download_url = "https://huggingface.co/MaziyarPanahi/gemma-2b-it-GGUF/resolve/main/gemma-2b-it.Q2_K.gguf"
     elif mv_selected_model == 'mistralai/Mistral-7B-Instruct-v0.2':
         lv_download_url = "https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q2_K.gguf"
 
@@ -160,21 +160,24 @@ def fn_generate_QnA_response(mv_selected_model, mv_user_question, lv_vector_stor
     if mv_selected_model == 'microsoft/phi-2':
         lv_model_path = "model/phi-2.Q2_K.gguf"
         lv_model_type = "pi"
+    elif mv_selected_model == 'google/gemma-2b':
+        lv_model_path = "model/gemma-2b-it.Q2_K.gguf"
+        lv_model_type = "gemma"
     elif mv_selected_model == 'mistralai/Mistral-7B-Instruct-v0.2':
         lv_model_path = "model/mistral-7b-instruct-v0.2.Q2_K.gguf"
         lv_model_type = "mistral"
-
-    print("Model Absolute location -" +lv_model_path)
+    
     
     print("Step4: Generating LLM response")
     fn_display_user_messages("Step4: Generating LLM response","Info", mv_processing_message)
 
-    lv_model = CTransformers(
-                                model="model/mistral-7b-instruct-v0.2.Q2_K.gguf",
-                                model_type=lv_model_type,
-                                max_new_tokens=1024,
-                                temperature=0.00
-                            )
+    lv_model = LlamaCpp(
+                            model_path=lv_model_path,
+                            temperature=0.75,
+                            max_tokens=2048,
+                            top_p=1,
+                            verbose=False
+                       )
     lv_retriever = lv_vector_store.as_retriever(search_kwargs={'k': 2})
     lv_qa_chain = RetrievalQA.from_chain_type(  llm=lv_model,
                                                 chain_type='stuff',
@@ -208,6 +211,7 @@ def main():
     mv_selected_model = col3.selectbox('Select Model',
                                         [
                                             'microsoft/phi-2',
+                                            'google/gemma-2b',
                                             'mistralai/Mistral-7B-Instruct-v0.2'
                                         ]
                                       )