Update app.py

app.py

@@ -1,22 +1,14 @@
-#pip install gradio
-
-
-#pip install datasets tqdm pandas matplotlib langchain sentence_transformers faiss-gpu langchain-community torch accelerate
-
 import gradio as gr
 import pandas as pd
-from tqdm.notebook import tqdm
-from datasets import Dataset
-import matplotlib.pyplot as plt
+from tqdm import tqdm
 from langchain.docstore.document import Document as LangchainDocument
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from sentence_transformers import SentenceTransformer
 from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
-from langchain.vectorstores import FAISS
+from langchain_community.vectorstores import FAISS
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores.utils import DistanceStrategy
 import torch
-
+import matplotlib.pyplot as plt
 
 # Set display option for pandas
 pd.set_option("display.max_colwidth", None)
@@ -29,12 +21,11 @@ with open("iplteams_info.txt", "r") as fp1:
 with open("match_summaries_sentences.txt", "r") as fp2:
     content2 = fp2.read()
 
-# Open and read the second file
+# Open and read the third file
 with open("formatted_playersinfo.txt", "r") as fp3:
     content3 = fp3.read()
 
-
-# Combine contents of both files, separated by three newlines
+# Combine contents of all files, separated by three newlines
 combined_content = content1 + "\n\n\n" + content2 + "\n\n\n" + content3
 
 # Split the combined content into sections
@@ -50,11 +41,7 @@ RAW_KNOWLEDGE_BASE = [
     for doc in tqdm(s)
 ]
 
-
-from langchain.text_splitter import RecursiveCharacterTextSplitter
-from sentence_transformers import SentenceTransformer
-from transformers import AutoTokenizer
-
+# Define markdown separators
 MARKDOWN_SEPARATORS = [
     "\n#{1,6}",
     "```\n",
@@ -87,11 +74,6 @@ fig.set_title("Histogram of Document Lengths")
 plt.title("Distribution")
 plt.show()
 
-
-from typing import Optional, List
-from langchain.text_splitter import RecursiveCharacterTextSplitter
-from transformers import AutoTokenizer
-
 EMBEDDING_MODEL_NAME = "thenlper/gte-small"
 
 def split_documents(
@@ -123,12 +105,6 @@ docs_processed = split_documents(512, RAW_KNOWLEDGE_BASE, tokenizer_name=EMBEDDI
 print(len(docs_processed))
 print(docs_processed[0:3])
 
-
-from langchain.vectorstores import FAISS
-from langchain_community.embeddings import HuggingFaceEmbeddings
-from langchain_community.vectorstores.utils import DistanceStrategy
-import torch
-
 print(torch.cuda.is_available())
 
 embedding_model = HuggingFaceEmbeddings(
@@ -144,10 +120,6 @@ KNOWLEDGE_VECTOR_DATABASE = FAISS.from_documents(
     distance_strategy=DistanceStrategy.COSINE,
 )
 
-
-import torch
-from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
-
 torch.random.manual_seed(0)
 
 model = AutoModelForCausalLM.from_pretrained(
@@ -171,7 +143,6 @@ generation_args = {
     "do_sample": False,
 }
 
-
 prompt_chat=[
     {
         "role":"system",
@@ -180,7 +151,6 @@ Give a comprehensive answer to the question.
 Respond only to the question asked , response should be concise and relevant to the question.
 provide the number of the source document when relevant.
 If the answer cannot be deduced from the context, do not give an answer""",
-
     },
     {
         "role":"user",
@@ -192,26 +162,24 @@ Question:{question}
         """,
     },
 ]
-RAG_PROMPT_TEMPLATE = tokenizer.apply_chat_template(
-    prompt_chat,tokenize = False,add_generation_prompt=True,
 
+RAG_PROMPT_TEMPLATE = tokenizer.apply_chat_template(
+    prompt_chat, tokenize=False, add_generation_prompt=True,
 )
 print(RAG_PROMPT_TEMPLATE)
 
 u_query = "give the match summary of royal challengers bengaluru and mumbai indians in 2024"
-# ret_text = KNOWLEDGE_VECTOR_DATABASE.similarity_search(query=u_query,k=3)
-retrieved_docs = KNOWLEDGE_VECTOR_DATABASE.similarity_search(query=u_query,k=3)
+retrieved_docs = KNOWLEDGE_VECTOR_DATABASE.similarity_search(query=u_query, k=3)
 
 context = retrieved_docs[0].page_content
 final_prompt = RAG_PROMPT_TEMPLATE.format(
-    question= u_query, context = context
+    question=u_query, context=context
 )
 
 output = pipe(final_prompt, **generation_args)
-print("YOUR QUESTION:\n",u_query,"\n")
-print("MICROSOFT 128K ANSWER: \n",output[0]['generated_text'])
+print("YOUR QUESTION:\n", u_query, "\n")
+print("MICROSOFT 128K ANSWER: \n", output[0]['generated_text'])
 
-# Define the function to handle queries
 def handle_query(question):
     retrieved_docs = KNOWLEDGE_VECTOR_DATABASE.similarity_search(query=question, k=3)
     context = retrieved_docs[0].page_content
@@ -221,8 +189,6 @@ def handle_query(question):
     output = pipe(final_prompt, **generation_args)
     return output[0]['generated_text']
 
-
-# Create a Gradio interface
 interface = gr.Interface(
     fn=handle_query,
     inputs="text",
@@ -231,5 +197,4 @@ interface = gr.Interface(
     description="Get the match summary of IPL teams based on your query.",
 )
 
-interface.launch(share=True)
-        
+interface.launch(sharing=True)