Update app.py

app.py

@@ -1,5 +1,4 @@
 import gradio as gr
-from llama_index.core.readers import SimpleDirectoryReader
 from llama_index.core import VectorStoreIndex, Document
 from llama_index.core.node_parser import SentenceSplitter
 from llama_index.core import Settings
@@ -8,14 +7,15 @@ from llama_index.llms.huggingface import HuggingFaceLLM
 import csv
 from docx import Document as DocxDocument
 import fitz
-
-# Define the list of LLMs with their names and models
+import os
+import torch
+os.environ['CUDA_LAUNCH_BLOCKING']="1"
+os.environ['PYTORCH_USE_CUDA_DSA'] = "1"
 lm_list = {
     "google/gemma-2-9b-it": "google/gemma-2-9b-it",
     "mistralai/Mistral-7B-Instruct-v0.3": "mistralai/Mistral-7B-Instruct-v0.3"
 }
 
-# Initialize the query engine globally
 query_engine = None
 
 def process_file(file):
@@ -48,13 +48,14 @@ def process_file(file):
 
 def handle_file_upload(file, llm_name):
     global query_engine
-
+    if torch.cuda.is_available():
+        torch.cuda.empty_cache()
     Settings.llm = HuggingFaceLLM(model_name=llm_name)
 
     documents = process_file(file)
 
     text_splitter = SentenceSplitter(chunk_size=512, chunk_overlap=10)
-    Settings.embed_model = HuggingFaceEmbedding(model_name="nomic-embed-text:latest")
+    Settings.embed_model = HuggingFaceEmbedding(model_name="nomic-ai/nomic-embed-text-v1.5", trust_remote_code=True)
     Settings.text_splitter = text_splitter
     index = VectorStoreIndex.from_documents(
         documents, transformations=[text_splitter], embed_model=Settings.embed_model
@@ -79,4 +80,4 @@ gr.Interface(
     title="Document Question Answering",
     description="Upload a document and choose a language model to get answers.",
     allow_flagging=False
-).launch()
+).launch()