Step-by-Step Guide to Building a RAG LLM App with LLamA2 and LLaMAindex

[@GerardoBarcia]

It would be amazing if you show us how to put all of this into production through an API! Thanks for your wonderful work! You Rock!!

[@timothylenaerts1123]

vLLM is easy enough to use, they provide a docker image, run that bad boy with whatever model you want and use their openAI endpoint then you can just use that in langchain or Llama index

[@Abhijit-VectoScalar]

@@timothylenaerts1123 wdym! could you please elaborate i actually want the same thing to do

[@sandeepmahale1941]

Krish, thank you for creating easy to understand videos

[@thomasmuller1521]

The libraries changed: from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt from langchain.embeddings.huggingface import HuggingFaceEmbeddings from langchain.embeddings import HuggingFaceEmbeddings from llama_index.embeddings.langchain import LangchainEmbedding import llama_index

[@AmitojSingh-tf9ex]

bro this line is giving error from llama_index.embeddings.langchain import LangchainEmbedding how do i find correct one

[@AmitojSingh-tf9ex]

how to make this thing running from langchain.embeddings.huggingface import HuggingFaceEmbeddings from langchain.embeddings import HuggingFaceEmbeddings from llama_index.embeddings import LangchainEmbedding embed_model=LangchainEmbedding( HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2"))

[@khyathinkadam5524]

while running import torch llm = HuggingFaceLLM( context_window=4096, max_new_tokens=256, generate_kwargs={"temperature": 0.0, "do_sample": False}, system_prompt=system_prompt, query_wrapper_prompt=query_wrapper_prompt, tokenizer_name="meta-llama/Llama-2-7b-chat-hf", model_name="meta-llama/Llama-2-7b-chat-hf", device_map="auto", # uncomment this if using CUDA to reduce memory usage model_kwargs={"torch_dtype": torch.float16 , "load_in_8bit":True} ) in colob i m getting import error stating that i need to install accelarate but i already have in in my env

[@nikhilanand9022]

I had one doubt is Vector Store Index use any embedding model behind it for creating the Index or how it create the embedding the vector store index

[@aravindsai2843]

Much awaited series, thank you krish Sir♥

[@Паша_Вамкон]

Thank you so much for this video!!! Very helpful!! I've managed to get a bit of understanding of LLM and to do my lab task!!!

[@chanishagarwal9103]

Thanks you krish for all your hard work. keep making such amazing videos.

[@smartwork7098]

Thanks man, it works well. (After correcting some changes made in llama and huggingface)

[@ahmadmasood3939]

I am facing problem while importing huggingfacellm. can you tell me what you did?

[@krvedhavakrishna7710]

@@ahmadmasood3939 while accessing tokens, in the create new token option, chnage token mode to write

[@koushikvenuganan3999]

hi, btw im also facing the same issues while importing llama and huggingface . Can u help me to fix the issues ?

[@MamadouMoussaBangoura]

thank you the cours i see in this rag topic , very good , i 'm french speaker but i understand your cours

[@Arkantosi]

Old tutorial. Most of the imports no longer work due to deprications.

[@HammieLicious]

SimpleInputPrompt is broken. still shows the concept tho

[@rubyrana7786]

Indeed great video. Please try to include the reason for using different approaches of doing any process like in earlier videos the model was loaded differently and here differently. A simple explanation of the reason behind using a specific approach can be useful for the beginners. As the approach changes when we move forward in more complex applications and different use cases.

[@y6bt2501]

Please make a video on RAG with CSV or database with local open source llm and with memory

[@atifsaeedkhan9207]

U r really a good instructor. ❤

[@user-fp3tm1jp2f]

Lets Goooooo

[@lixiasong3459]

Thank you, Sir, you are amazing!

[@user-hj9ck9rp4c]

I am getting the error of VectorStorIndex from LLama_index

[@ByYouTube2]

use- from llama_index.core import VectorStoreIndex,SimpleDirectoryReader,ServiceContext

[@akshatapalsule2940]

Thankyou so much Krish, it was worth the wait :)

[@RanjitSingh-rq1qx]

Sir everything fine, but you have missed just one thing in this project, that why you have implemented or build prompt and why you have not used that prompt, why you have gone with default prompt ? And remaining part was so good with good explanation ❤️

[@krishnaik06]

More examples i will come up with...this basic to intermediate RAG system

[@MLAlgoTrader]

You are amazing.

[@GaneshEswar]

waiting for next video, please upload it ...

[@nunoalexandre6408]

Love it

[@ivanrowland142]

Love this. When is the next instalment?

[@stevefisher35]

Thanks for the detailed run through, very useful . One question I have is on the two PDF documents you have. Are these available anywhere, just for testing purposes?

[@sravan160]

i have some doubts in implementing the code.,can u help?

[@eswararya196]

Yeah ask me?

[@DoomsdayDatabase]

Hi krish sir! they have updated the service_context to settings.llm and i am not able to understand how to implement it into this code. Please help! Thanks!

[@amritsubramanian8384]

gr8 video

[@rajarams3722]

Sorry Krish...you started the playlist saying you would explain the difference in llamaindex vs langchain and why we should use either and when...but this play list with three videos does not deal with that, instead explains some hands on....Thanks for your work, but I think the assurances are not met..The play list has not been updated since Jan 24 ?

[@rafiquemohammed3029]

Hey @krishnaik06 Please create video on Intent Classification using LLM and VectorDB. How to create and train Intents and use to ask Custom Question without LLM?

[@charmilagiri4602]

Sir Instead of using LLamA2 model from huggingface, Can we try the quantized llama model? If we use the quantized llama model will the output accuracy varies?

[@miteshgarg9420]

Hey Krish, amazing video again. Can you please help to create a similar solution for custom text 2 sql

[@hassubalti7814]

sir greet method to teach us as well as gaining good grip on english. please make video about token used in llama 2 model used

[@MK5491]

@krishnaik06 Sir Thank you for this knowledgeable video, My question is which Evalution model we should use to show the accuracy in terms of answer and context retrieval. If possible, will you please create one video on evaluation method for RAG application.

[@akashchavan3353]

@krishnaik06 sir I also want this thing, can you please create one video on evolution method. Thanks

[@anant9421]

@krishnaik06 yes please can you please create a video on evaluation method for RAG application

[@yashghugardare519]

Sir..instead of RAG with pdf..make a video on Rag with videos.. which will process videos and be able to answer questions based on the video

[@IndrajeetKhatri-y6z]

Can you please upload the video for using the MistralAI model....

[@tarekferradji]

Can I combine the LORA fint tuning for example and a RAG FOR THIS llm, can this work give me very interesting performances?

[@yuvrajthakur5728]

Hey Krish, why you left PWskills masters in data science course. I joined this because of you. But I am seeing there new tutors. I joined this course because of you only.

[@dillikaextrovert]

Hello Krish The list of accessories you have mentioned, is not having right links for Amazon. Can you please give me the link for the writing pad which you use ?

[@chinnibngrm272]

Sir as I am a student.. I am not having gpu in my machine... I am not able to do projects with this open source llms n also with open ai... Can u please help us to solve the resource errors by using other models

[@atifsaeedkhan9207]

Is it possible to use local llms instead huggingface directly? I have ollama nd lmstudio installed.

[@rubyrana7786]

Why did we use a separate Embedding model here while in the earlier video of this playlist we directly used VectorStoreIndex on the documents. So why did we follow different approaches while creating similar Applications? Is it because of the different Model or it is just a different approach and can be done either way ?

[@nikhilanand9022]

I had one doubt is Vector Store Index use any embedding model behind it for creating the Index or how it create the embedding the vector store index

[@Glimmer-t44]

@@nikhilanand9022 By default, VectorStoreIndex use OpenAI Embeddings if no other embeddings are explicitly specified.

[@shreyavalte3077]

I have a very big question

[@yetanotheremail]

Han please ask

[@minimal2224]

Lol the suspense is killing me

[@Lemonboi-lmn21]

any updates ?

[@blakdronzer]

there is an issue downloading the model LLAMA2 / LLAMA3 - as it requires to be authenticated / accepted by their team. Even after getting it authenticated / accepted - when we try download the code by your pattern / huggingface_clt - it still says - access denied to the config file - and hence cannot download any model locally - so if you could practically share insights or build a new video that pertains to the latest situation, will be awesome

[@rajarams3722]

Please try in google colab..I tried in my local laptop and many things failed.

[@blakdronzer]

Well i discovered lately that this models are available only for pro accounta. Since I subscribed to one monther, it is working fine now. Strange... Marking it open but available only for subscribers

[@goldy5553]

Library is pretty messed up, nothing is working everwhere there is a module import error and function is missing or deprecated. if you found this, don't worry guys we are on same page. Sir could you please check if there is a some issues or what they have done to library

[@eswararya196]

If you are having moldule import error then use llama_index.core Instead of llama_index

[@madhujegishetti4102]

ImportError: cannot import name 'VectorStoreIndex' from 'llama_index' (unknown location)--getting this error

[@jeruc.r7382]

Same error 😢 did you cleard that

[@user-dc8hw3hz5e]

from llama_index.core import VectorStoreIndex,SimpleDirectoryReader,ServiceContext

[@vinayaktiwari4463]

Hi Krish , which vector store have you utilised here? there was no mention of such in the code

[@VivekGuptaMusic]

not saving the embeddings in any vector store directly using it

[@sayanghosh6996]

06:00 !pip install -q llama-index llama-index-llms-huggingface from llama_index.core import VectorStoreIndex,SimpleDirectoryReader,ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt

[@yourealive1]

thank you so much

[@Glimmer-t44]

"ServiceContext" is now deprecated and replaced with "Settings" object. You may need to update this tutorial. Also, at 16:19 why did you used two different embeddings? HuggingFaceEmbeddings and LangChainEmbedding? Isn't just any one is enough?

[@rajarams3722]

Yes, there are lots of gaps such as these in his recent videos..Seems like he is in a hurry to have his hands on every topic that emerges and he is losing direction..starting new playlist and leaving it to start new ones..

[@achukisaini2797]

How to reduce hallucination? if answer is not in context then it is hallucinating .

[@soulfuljourney22]

may be you can modify the prompt to answer for not in context situation

[@IamMarcusTurner]

literally prompt the LLM if not in document tell the LLM to say it does not know.

[@riteshsingh811]

If the content is present , still its hallucinating , there are certain advance RAG techniques like Window Sentence Retrieval and Auto Merging Retrieval that can help.It will help in improving the context. Just try read regarding it and implement. It will help u. Also tuning agent to not give answer when it doesn't know helps in case of unknown scenario.

[@SagarGuhe]

Apparently imports from llama_index are not working as of now.

[@vishalpawar4150]

Can we please implement this using TS/JS ?

[@nostalgia18rishi]

I can't login using hugging face cli on colab. I have pasted the token from hugging face and control enter. The cell on my collab keeps running.

[@Narutome30]

Y is he using google colab rather than vs code And also please answer this question -> can we use vs code to run seamless m4t meta model

[@vivekshindeVivekShinde]

I have lots of pdf documents data and want to create a custom chatbot based on it. Then which one will be better: Langchain or Llamaindex?

[@RanjitSingh-rq1qx]

Llamaindex for indexing purpose, and langchain used for response of query with Prompt by the langchain LLM, and used Gemini pro as a LLM model. Will be great combination of all these technologies ❤

[@vivekshindeVivekShinde]

@@RanjitSingh-rq1qx Thanks for suggestions. I am looking for Open Source. So while Indexing in the Llamaindex, it doesn't use OpenAI api or something right?

[@RanjitSingh-rq1qx]

@@vivekshindeVivekShinde yes all are open source

[@chinnibngrm272]

Guys can you please share the implementation of this by mixing llama index, Langchain, Gemini pro.... Please ... It will be very helpful 😊😊

[@nikhilanand9022]

@@vivekshindeVivekShinde I think when we use the Vector Store index it use the openai embedding model api for creating the index can you please confirm once ?

[@lesstalkeatmore9441]

i am getting error while i am import library

[@MatkoZaja]

Is there a way to ensure that once PDFs are processed, they do not need to be reprocessed every time the script runs, but rather that a cached database can be stored? Does anyone have code for this?

[@chinnibngrm272]

I am getting Runtimeerror: CUDA Error.... While running index= VectorStoreIndex.from_documents (docs, service_context=service_context) Sir please provide a solution to run with cpu....

[@sandhya77331]

how to integrate streamlit for front-end into this

[@shehrozkhan9563]

Can we add in conversation history to this app?

[@fadhilayosof5927]

can you make a video to create flowchart by LLM

[@aritrasaha1854]

how to do it if I have a csv file instead of a pdf

[@Playstore-zc5xk]

How to convert this into end to end?

[@ridj41]

where to get the data from, like you have done in this case? \

[@himanshudeswal9895]

Hi, can anyone tell me how to download these raw pdf’s for hands on please??

[@abhinandansharma3983]

where i will get this dataset

[@harshab2743]

i am getting an error while importing vectorstoreIndex from llamaIndex saying that llamaIndex doesn't exist. can someone help

[@ShreyasR-vr1es]

Trying using the import like this instead: from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt This should work for you!

[@aiwithuday]

@@ShreyasR-vr1es also pip install llama-index-llms-huggingface

[@user-bb5nz6cf5j]

Sir , the llama index library is modifying everyday , and there are many import errors in the code , can you tell me the suitable version of llama-index to run the code

[@ishratsyed2857]

I was having the same issue, I tried installing version 0.9.40 and it's working now

[@sebastienmaillet9371]

@@ishratsyed2857 i tried to install llama_index version 0.9.40 but i got the following message: ImportError Traceback (most recent call last) in () ----> 1 from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext 2 from llama_index.llms import HuggingFaceLLM 3 from llama_index.prompts.prompts import SimpleInputPrompt ImportError: cannot import name 'VectorStoreIndex' from 'llama_index' (unknown location)

[@sebastienmaillet9371]

do you know what i might be missing ?

[@santhoshmanoharan8969]

@@sebastienmaillet9371 I have tried the same code from my local anaconda environment, I'm getting error with the importing packages but it is working fine when I use Google Colab, can anyone explain why?

[@ShreyasR-vr1es]

Trying using the import like this instead: from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt This should work for you!

[@sumanmaity3162]

Hello Krish, I'm getting a basic error as below. Can you please help? ImportError Traceback (most recent call last) in () ----> 1 from llama_index import VectorStoreIndex,SimpleDirectoryReader,ServiceContext 2 from llama_index.llms import HuggingFaceLLM 3 from llama_index.prompts.prompts import SimpleInputPrompt ImportError: cannot import name 'VectorStoreIndex' from 'llama_index' (unknown location)

[@ShreyasR-vr1es]

Trying using the import like this instead: from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt This should work for you!

[@sumanmaity3162]

@@ShreyasR-vr1es Thank you, other 2 worked but I'm getting error in below ModuleNotFoundError Traceback (most recent call last) in () ----> 1 from llama_index.llms.huggingface import HuggingFaceLLM ModuleNotFoundError: No module named 'llama_index.llms.huggingface'

[@sumanmaity3162]

Please ignore, it worked, had some installation issues. Thank you so much.

[@darshitshah8668]

@@sumanmaity3162 I am facing same error, how did it get resolved for you?

[@sumanmaity3162]

@@darshitshah8668 Please reinstall, it should work

[@user-gk7ox3of4b]

i need those pdfs

[@KumR]

Hey Krish. Video is cool. But can you tell us how we will know what are the different things we will need to import . You may have done lot of research. Kindly point us to the source of truth.

[@khyathinkadam5524]

while running import torch llm = HuggingFaceLLM( context_window=4096, max_new_tokens=256, generate_kwargs={"temperature": 0.0, "do_sample": False}, system_prompt=system_prompt, query_wrapper_prompt=query_wrapper_prompt, tokenizer_name="meta-llama/Llama-2-7b-chat-hf", model_name="meta-llama/Llama-2-7b-chat-hf", device_map="auto", # uncomment this if using CUDA to reduce memory usage model_kwargs={"torch_dtype": torch.float16 , "load_in_8bit":True} ) in colob i m getting import error stating that i need to install accelarate but i already have in in my env

[@AnandYadav-gv1xw]

from llama_index import VectorStoreIndex, SimpleDirectoryReader, ServiceContext ImportError: cannot import name 'VectorStoreIndex' from 'llama_index' (unknown location)

[@Sohammhatre10]

Simple directory reader too

[@ShreyasR-vr1es]

Trying using the import like this instead: from llama_index.core import VectorStoreIndex, SimpleDirectoryReader, ServiceContext from llama_index.llms.huggingface import HuggingFaceLLM from llama_index.core.prompts.prompts import SimpleInputPrompt This should work for you!

[@arnavdeshmukh2820]

facing issue index=VectorStoreIndex.from_documents(documents,service_context=service_context) can anyone help