- Видео 24
- Просмотров 5 943
M&M Tech
Франция
Добавлен 10 мар 2024
Welcome to M&M Tech 🌐 Join our community for clear explanations, insightful discussions, and a journey towards tech mastery. Subscribe and become a tech enthusiast! 💬💻
Multimodal RAG Full Tutorial | Langchain LCEL & LM Studio
This video is a complete step by step guide to multimodal rag.
With this tutorial, you will learn how to build a full end-to-end Multimodal RAG (Retrieval Augmented Generation) pipeline in Python that can let you chat with any document using the power of generative AI to capture both its visual and textual contents.
Specifically, you will learn how to deploy your multimodal LLMs locally using LM Studio and how to build end-to-end ingestion and inference pipelines using Langchain LCEL.
The topics covered are :
- Pipeline design for document processing and question answering
- Local deployment of LLMs, MLLMs and embedding models with LM Studio
- Layout Analysis and OCR with Unstructured.io
- Vect...
With this tutorial, you will learn how to build a full end-to-end Multimodal RAG (Retrieval Augmented Generation) pipeline in Python that can let you chat with any document using the power of generative AI to capture both its visual and textual contents.
Specifically, you will learn how to deploy your multimodal LLMs locally using LM Studio and how to build end-to-end ingestion and inference pipelines using Langchain LCEL.
The topics covered are :
- Pipeline design for document processing and question answering
- Local deployment of LLMs, MLLMs and embedding models with LM Studio
- Layout Analysis and OCR with Unstructured.io
- Vect...
Просмотров: 358
Видео
Multimodal RAG in Single LCEL Pipeline | Langchain Multimodal RAG #EP4
Просмотров 82Месяц назад
This video is the last out of four episodes of a series dedicated to multimodal RAG. In this third episode, you will learn the following : - How to build the inference pipeline for multimodal RAG - How to use the Langchain LCEL chaining and pipe logic - How to use the LCEL runnable passthrough and assign method - How to use the LCEL runnable branch for conditional chaining - How to run LCEL cha...
Prompts & LCEL Chains for Retrieval & Question Answering | LangChain Multimodal RAG EP#3
Просмотров 81Месяц назад
This video is the third out of four episodes of a series dedicated to multimodal RAG. In this third episode, you will learn the following : - How to build the inference pipeline for multimodal RAG - How to build simple chains using the Langchain LCEL framework - How to write robust prompts for textual/visual question answering tasks. Don't forget to like and subscribe to the channel if you appr...
Image Description & Local Vectorstore | Langchain LCEL + Qdrant | Multimodal RAG #EP2
Просмотров 184Месяц назад
This video is the second out of four episodes of a series dedicated to multimodal RAG. In this first episode, you will learn the following : - How to set up your document processing pipeline for multimodal RAG - How to build chains with langchain LCEL - How to use multimodal LLMs for image description - How to build a Qdrant vectorstore and upsert documents to it Don't forget to like and subscr...
Understand Multimodal RAG Pipelines & Deploy MLLMs on LM Studio
Просмотров 3842 месяца назад
This video is the first out of four episodes of a series dedicated to multimodal RAG. In this first episode, you will learn the following : - What is multimodal RAG - How to set up your document processing pipeline for multimodal RAG - Two different ways to set up your question answering pipeline for multimodal RAG - How to deploy multimodal LLMs locally on CPU with LM Studio Further reading ma...
Qdrant Retrieval With Local Embeddings on LM Studio & Ngrok | RAG Tutorial
Просмотров 5333 месяца назад
Welcome back to the channel for this new RAG tutorial ! TLDR : This video is a retrieval tutorial where you will learn to use Qdrant as a Vector Store after deploying your local embedding model with LM Studio and Ngrok. In this video, you will learn how to use LM Studio to deploy virtually any HuggingFace embedding model locally without the need for GPUs or excessively performant CPUs and RAM. ...
Deploy LLMs Locally On CPU With LM Studio & LangChain
Просмотров 1 тыс.3 месяца назад
In this video, you will learn how to use LM Studio to deploy virtually any HuggingFace LLM locally without the need for GPUs or excessively performant CPUs and RAM. With LM Studio, you can deploy opensource LLMs such as Gemma, Mistral, Phi and Llama as local API servers similars to the OpenAI API. This allows for seamless integration with frameworks such as Langchain and LlamaIndex by simply tr...
How To Get Started With DBT (Data Build Tool) | Full Step By Step Tutorial
Просмотров 1104 месяца назад
This video is a step-by-step tutorial on how to get started with dbt (Data Build Tool). You will learn what dbt is, explore use case examples, and understand how dbt works. The tutorial includes a detailed demo of how to install dbt Core on both Windows and Mac, how to install the database plugin (using Google BigQuery in the demo), and how to initialize a dbt project. All of this is done withi...
Ultimate RAG Tool for Document OCR & Extraction - UnstructuredIO Tutorial
Просмотров 6654 месяца назад
In this video you will learn step by step how to use the Unstructured.io Serverless API for document OCR and extraction with a FREE api key. If you're wondering how to extract text from pdf files (or any other file type such as excel sheets word or PowerPoint slides) through advanced layout analysis and optical character recognition, and specifically how to also extract tables, images, formulas...
Jupyter Notebook Virtual Environment Setup | Python Tutorial
Просмотров 664 месяца назад
In this video, you will learn how to create virtual environments with a specific version of python and launch jupyter notebooks within your newly created environments. You will specifically learn about the link between python environments and jupyter kernel and how to use the ipykernel package and jupyter kernelspec commands to establish this link. You can read more on virtual environments in t...
Langchain LCEL Intuitively Explained + Notebook Examples | Langchain Tutorial
Просмотров 1466 месяцев назад
In this video, I provide an intuitive explanation of LangChain Expression Language (LCEL). You'll discover how to create various types of chains in LCEL, gaining a comprehensive understanding of this powerful tool. Timestamps : 00:00 Intro 00:39 Runnables & Pipe Operator 02:04 Demo - Classic example 03:14 Demo - Create custom runnable 04:30 Demo - Multiple input chain 06:53 Demo - Concurrent ch...
Multi-language Data Anynomization with Langchain & Presidio - Full Step By Step Tutorial
Просмотров 2706 месяцев назад
In this video you will learn step by step how to use Microsoft Presidio and LangChain to anonymize sensitive data in different languages, and how to integrate this step with GPT. 🔔 LIKE, SUBSCRIBE AND LET'S DISCUSS IN THE COMMENTS :) !! Pip command for installation : !pip install upgrade langchain langchain-openai langchain-experimental presidio-analyzer presidio-anonymizer spacy Faker langdete...
Data Anonymization With Microsoft Presidio - Full Step By Step Tutorial
Просмотров 8966 месяцев назад
Data Anonymization With Microsoft Presidio - Full Step By Step Tutorial
Spark Tutorial | Environment Setup in 5 Minutes | Full Step By Step Guide | Learn Apache Spark
Просмотров 1187 месяцев назад
Spark Tutorial | Environment Setup in 5 Minutes | Full Step By Step Guide | Learn Apache Spark
GCP Tutorial | Google Cloud Storage API With Python | Full Step By Step Guide | Learn GCP
Просмотров 1828 месяцев назад
GCP Tutorial | Google Cloud Storage API With Python | Full Step By Step Guide | Learn GCP
Google Sheets Python Tutorial | Read & Write With Google Sheets API
Просмотров 1688 месяцев назад
Google Sheets Python Tutorial | Read & Write With Google Sheets API
GCP Tutorial | How To Add Identity-Aware Proxy (IAP) to App Engine | Learn GCP
Просмотров 568 месяцев назад
GCP Tutorial | How To Add Identity-Aware Proxy (IAP) to App Engine | Learn GCP
Python Multithreading vs Multiprocessing vs Asyncio Explained in 10 Minutes | Learn Python
Просмотров 1548 месяцев назад
Python Multithreading vs Multiprocessing vs Asyncio Explained in 10 Minutes | Learn Python
GCP Tutorial | Deploy Python Flask App on Google App Engine | Learn GCP With Python
Просмотров 2319 месяцев назад
GCP Tutorial | Deploy Python Flask App on Google App Engine | Learn GCP With Python
GCP Tutorial | Create GCP Account & Install gcloud SDK | Learn GCP
Просмотров 559 месяцев назад
GCP Tutorial | Create GCP Account & Install gcloud SDK | Learn GCP
GCP Tutorial | Create Virtual Machine & Connect To It | Learn Google Compute Engine
Просмотров 399 месяцев назад
GCP Tutorial | Create Virtual Machine & Connect To It | Learn Google Compute Engine
Amaazing Tutorial. One thing to add: The RAG examples you use have integer type IDs, then in Step 4, when creating the embeddings, that raises an error, because the IDs must be String UUID. Simply change the document IDs with UUIDs and voila!
Thanks for the feedback ! The fix will soon be added to the notebook ! 😉
Very informative and helpful! Can we connect in twitter or linkedin?
Thanks ! And yes the channel's handle is @MMTech24 on twitter 😉
Could you please do the part where you said we can find out from which page, document, paragraph etc.. we got our answer from. I really loved this video thank you!
Thank you ! The rag-with-sources tutorial will indeed be the next video on the channel, coming very soon ! 😉
@ when will you upload that
Should be up by the end of the week
And as usual, here's the link to the notebook : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Multimodal.ipynb If you have any problem following the tutorial or any suggestion for future videos, the comment section is all yours !
And as usual, here's the link to the notebook : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Multimodal.ipynb If you have any problem following the tutorial or any suggestion for future videos, the comment section is all yours !
Can we connect to Postgresql database ?
Hey ! In which context are you trying to connect to a postgresql database ? Could you explain your use case?
@@MnMTech889 I want to run LLM locally with RAG technic which connect to my organization database then query data from it.
Oh then you're looking for pgvecto.rs (the rust version of pgvector which has better features). It has langchain wrappers too ! I will be making a pgvecto.rs + postgresql with sqlalchemy tutorial soon.
@@MnMTech889 thank you very much. I am confusing how to run pgvector.py with LM Studio windows desktop version ?
Main things explained in little time. I love this method. Thank you
Glad it was helpful !
Full notebook available on our github page : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Multimodal.ipynb If you have any suggestions for future videos, the comment section is all yours !
Your content has been really helpful to me. Keep up with the good work, and I look forward for the next episode.
Thanks ! Will make more and more specialized content, don't hesitate to point us to a specific topic of interest to you 😉
And as usual, here's the link to the notebook : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Multimodal.ipynb If you have any problem following the tutorial or any suggestion for future videos, the comment section is all yours !
what would be the purpose of a reranker in this context?
With a large vectostore (thousands to millions of vectors), the dense retriever can very quickly bring it down to the 100 or so best candidates for a query. A reranker (much slower but much more accurate) would then reorder the 100 results.
We will be doing a series dedicated to improving retrieval very soon !
And as usual, here's the link to the notebook : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Multimodal.ipynb If you have any problem following the tutorial or any suggestion for future videos, the comment section is all yours !
is the notebook available?
Was planning on pushing it at the end of the series but you're right it makes no difference. Just give me a sec... done !
👍
👍
Could you make a video on multimodal rag?
definitely planning on making one 😉 it'll most likely be next
Waiting
And as usual, here's the link to the notebook : github.com/MnMTech-hub/tutorials/blob/master/LM-Studio/Local-Embedding.ipynb If you have any problem following the tutorial or any suggestion for future videos, the comment section is yours !
I'm getting connection error..
At what point ?
I just followed the tutorial and at the end while running "stream("prompt")" I'm like getting connection error. Btw, I'm doing it in Google colab
Though,the model is generating responses in LM studio but in notebook giving no result
Oh I see you're trying to access it as a public url, you need ngrok to do this with colab. Try this : medium.com/@aivaa.io.team/how-to-expose-your-local-llm-to-the-internet-using-ngrok-and-aivaa-1bbc92074987
Very good point indeed, I'll add this in the description and in future videos about local LLMs
Can we integrate this notebook into our projects? Like, creating API endpoints using flask and integrate into our web apps?
yes Indeed, we intend to make tutorials on this with LangServe (fastapi integration) and also further tutorials on how to deploy with Ollama or directly with LlamaCPP without going through LM Studio
Great! Subscribed..
And here's the notebook for this tutorial : github.com/MnMTech-hub/tutorials/tree/master/LM-Studio Feel free to make any suggestion for future videos in the comment section 😉
The full notebook for this tutorial can be found in this github repo : github.com/MnMTech-hub/tutorials/tree/master/Unstructured-io We'll make sure to upload all code and material from the channel to this repo (for previous as well as future videos) so stay tuned !
Can you share the github link or the code in the comments section? Would like to try it out
Great idea ! We'll soon make a github page with all the channel's tutorial notebooks and material 👌
Could you do a detailed video about this in RAG? For example lets say I have several documents which have employee ID number that I want to anonymize before I create embeddings and also anonymize during the Q/A. How could I create a custom fake data for these employee ID numbers cuz I looked at langchains example using polish numbers but is it possible to create custom faker class for employee ID or anything not built in already in faker?
Will definitely do more tutorials involving RAG and agents in the upcoming weeks. For custom fakers on employee ID it should be possible with an appropriate regex + faker with presidio.
Hello.. I just found your channel and I'm very grateful about these langchain videos. I've never ever understood LCEL before no matter who I watched and I honestly did from this video. It made so much sense to me and please please please could you do a video on all of the different runnables that LCEL has and the difference between them? Thank you so much!
Thank you ! Many more LCEL tutorials are on the way detailing each runnable, stay tuned !
incredibly helpful! thank you :)
you're welcome :)
Perfect
👏👏👏👏
What can I do if the Spacy Model is not good enough for anonymization?
You can add a custom transformers model as your engine instead of Spacy, here is the documentation link for that : microsoft.github.io/presidio/samples/python/transformers_recognizer/ Also, here you have code samples for different ways to use and customize presidio : microsoft.github.io/presidio/samples/
@@MnMTech889 Thanks! I have solved it now with Flair
great content !
amazing !
Keep it going
Amazing
Good explaining thank you
You're welcome 😇
thats a dirty url
why so?
Keep it up
Nice video !