- 103
- 36 705
Aritra Sen
India
เข้าร่วมเมื่อ 5 มิ.ย. 2009
Hello everyone. This is Aritra Sen. I am a Data Scientist by profession with more than 14 years of experience in this fields. I am exploring more about Data Science/Machine Learning is my passion. I am here to share my knowledge and contribute to Data Science community to learn and grow together with you all.
Join this WhatsApp community for updates chat.whatsapp.com/EAXHrY5lmSBJJfWMfNyCMd
This channel will cover topics like #datascience #machinelearning #python #generativeai #llm etc. More contents to follow.
Join this WhatsApp community for updates chat.whatsapp.com/EAXHrY5lmSBJJfWMfNyCMd
This channel will cover topics like #datascience #machinelearning #python #generativeai #llm etc. More contents to follow.
LangGraph: Build LLM based SQL Database Agents using LangGraph : Part 6
Notebook : github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/SQL_Database_Agent_LangGraph.ipynb
มุมมอง: 156
วีดีโอ
LangGraph : Bring humans in the loop with LLM Agents : Part5
มุมมอง 32214 วันที่ผ่านมา
Notebook link: github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/Bring_Human_In_the_loop_Agents.ipynb
Deep Dive into Hybrid search techniques in RAG: BM25, MMR, Reciprocal Ranking, Metadata filtering
มุมมอง 31628 วันที่ผ่านมา
Hybrid Search: BM25Retriever , MMR , Reciprocal Ranking Metadata Filtering: Metadata Extraction using LLM and SelfQueryRetriever Notebooks present in this github repo: github.com/aritrasen87/LLM
Deep Dive into Large Language Models : From RNN to RLHF
มุมมอง 660หลายเดือนก่อน
0:00 Introduction 4:40 RNN and it's issues 32:05 LSTM and how its solves the issues of RNN 46:54 Sequence2Sequence Models (Encoder-Decoder) 54:13 Bidectional RNN with Attention Mechanisam 1:07:37 Self Attention,Multihead Attention, Masked Multihead and Sliding Window Attention 1:32:15 Positional Encodings 1:43:00 Grouped Query Attention 1:54:00 KV Cache 2:05:48 Rolling Buffer Cache 2:11:55 Pref...
LangGraph : Self Corrective RAG (CRAG) to Reduce hallucination in RAG Pipeline : Part4
มุมมอง 585หลายเดือนก่อน
Learn how you can use LangGraph to create corrective RAG to reduce hallucination Notebook : github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LangGraph_04_Self_Corrective_RAG.ipynb
Deep dive into RAG Chunking Strategies : CharacterText Splitter to Semantic Chunking
มุมมอง 389หลายเดือนก่อน
- Why do we need chunking in RAG - Different chunking strategies and Pros and Cons - CharacterText Splitter - RecursiveCharacter Text Splitter - TokenText Splitter - based embedding max tokens - LLM Context length - Semantic Chunking Notebook: github.com/aritrasen87/LLM/blob/main/Chunking_Strategy_DeepDive.ipynb
Deep dive into LangChain Expression Language (LCEL): Pipe Operator , Runnables, Async & Streaming
มุมมอง 4362 หลายเดือนก่อน
Topics covered: - Pipe Operator - Understanding runnables - RunnableParallel - RunnablePassthrough - RunnableLambda - Assign - Performance improvement (inference speed) - Async invoke - Batch support - Async Batch - Using Itemgetter with LCEL - Bind tools - Stream Support Notebook : github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LCEL_DeepDive.ipynb
LangGraph: Create Business Intelligence tool using Multi Agent Supervisor : Part3
มุมมอง 1.8K2 หลายเดือนก่อน
This video talks about how you can use Multi Agent Supervisor to retrieve data from Internet/RAG(Using Search or RAG Tool) and create visualization using Python Coder Agent. Notebook: github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LangGraph_03_MultiAgent_With_Supervisor.ipynb
LangGraph: Understanding OpenAI Tool Calling and Integration of Tools with LangGraph : Part2
มุมมอง 7462 หลายเดือนก่อน
- Understanding OpenAI tool invocation - Integration of Tool calling with LangGraph Notebook: github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LangGraph_02_ToolCalling.ipynb
LangGraph: Getting Started: Step by Step tutorial to build Agents : Part1
มุมมอง 3.2K2 หลายเดือนก่อน
- Nodes - Edges - Graph creation - Conditional Edges - Adding LLM calls - Adding RAG Calls Notebook: github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/LangGraph_01_GettingStarted.ipynb
Meta Llama3 from concepts to hands-on coding
มุมมอง 1363 หลายเดือนก่อน
Notebook link: github.com/aritrasen87/LLM/blob/main/LLama3.ipynb
Better RAG with MultiIndexRetriever : Retrieve full documents
มุมมอง 1953 หลายเดือนก่อน
Notebook link: github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/MultiIndexRetriever.ipynb
Better RAG with HyDE - Hypothetical Document Embedding
มุมมอง 4733 หลายเดือนก่อน
Notebook code : github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/HyDE.ipynb
Better RAG with Query Re-Writing and Document Re-Ranking with Cross Encoder
มุมมอง 5423 หลายเดือนก่อน
Code notebook : github.com/aritrasen87/LLM_RAG_Model_Deployment/blob/main/cross_encoder.ipynb Langchain playlist : th-cam.com/play/PLOrU905yPYXItzOax1OUsgkehvlM7wIK5.html&si=IrGMaeIjqnWBsW2I
Better LLM Architecture with Semantic Router and Function Calling
มุมมอง 5784 หลายเดือนก่อน
Github Repo for notebooks: github.com/aritrasen87/semantic_routers Join my Data Science Focused WhatsApp channel : chat.whatsapp.com/EAXHrY5lmSBJJfWMfNyCMd Creator of Semantic Router: www.aurelio.ai/semantic-router
EP8: RAG LLM App: Integrating everything and making inference
มุมมอง 5115 หลายเดือนก่อน
EP8: RAG LLM App: Integrating everything and making inference
EP7: RAG LLM App : How to build front end Demo app with Gradio
มุมมอง 4885 หลายเดือนก่อน
EP7: RAG LLM App : How to build front end Demo app with Gradio
EP6: RAG LLM App : Build Endpoints with FastAPI and add Pydantic Validation
มุมมอง 5035 หลายเดือนก่อน
EP6: RAG LLM App : Build Endpoints with FastAPI and add Pydantic Validation
EP5: RAG LLM App: How to Write Modularized code and Build RAG
มุมมอง 3365 หลายเดือนก่อน
EP5: RAG LLM App: How to Write Modularized code and Build RAG
EP4: RAG LLM App: Load LLMs with CTransformer and TogetherAPI
มุมมอง 3895 หลายเดือนก่อน
EP4: RAG LLM App: Load LLMs with CTransformer and TogetherAPI
EP3: RAG LLM App: How to manage secrets with dotEnvFile
มุมมอง 3855 หลายเดือนก่อน
EP3: RAG LLM App: How to manage secrets with dotEnvFile
EP2: RAG LLM App : Managing Environment and Dependencies
มุมมอง 5385 หลายเดือนก่อน
EP2: RAG LLM App : Managing Environment and Dependencies
EP1: RAG LLM App with FASTAPI and Gradio: Introduction
มุมมอง 1.9K5 หลายเดือนก่อน
EP1: RAG LLM App with FASTAPI and Gradio: Introduction
Mistral Spelled Out: Sparse Mixture of Experts (MoE) : Part 10
มุมมอง 776 หลายเดือนก่อน
Mistral Spelled Out: Sparse Mixture of Experts (MoE) : Part 10
Mistral Spelled Out: Grouped Query Attention : Part 8
มุมมอง 1776 หลายเดือนก่อน
Mistral Spelled Out: Grouped Query Attention : Part 8
Mistral Spelled Out: Prefill and Chunking : Part 9
มุมมอง 2626 หลายเดือนก่อน
Mistral Spelled Out: Prefill and Chunking : Part 9
Mistral Spelled Out: Rolling Buffer Cache : Part7
มุมมอง 2456 หลายเดือนก่อน
Mistral Spelled Out: Rolling Buffer Cache : Part7
Mistral Spelled Out : KV Cache : Part 6
มุมมอง 3316 หลายเดือนก่อน
Mistral Spelled Out : KV Cache : Part 6
Mistral Spelled Out : RMS Norm : Part 5
มุมมอง 3026 หลายเดือนก่อน
Mistral Spelled Out : RMS Norm : Part 5
Mistral Spelled Out : Rotary Positional Encoding : Part 4
มุมมอง 1816 หลายเดือนก่อน
Mistral Spelled Out : Rotary Positional Encoding : Part 4
Can you also show how to stream the langgraph output from final node
App.stream should work, is it not working?
@@AritraSen I made an agent using langgraph it streams node wise
@Aritra Sen, All your LangGraph sessions are too good. Please provide One real time use case end to end project on this please.
you can also invoke if you write chain = RunnablePassthrough() | RunnableLambda(convert_to_capital) | RunnablePassthrough() | prompt 😉
If I want to achieve the same thing with anthropic claude I have created my own parser, but when passing messages for supervisor prompt there is list of human messages and gives error conversation should be between user and assistant
Hey sorry I have not used Claude models and don't have the access also... Sorry won't be able to help on this..
Bhai u used the same code as: - th-cam.com/video/3w_D1L0F-uE/w-d-xo.html atlease credit to de diya kar bhai !!
Not sure about that video.. But my content was highly influenced by the chromadb deeplearning.ai short course for this video... That I can tell you...
Hi, Well put explanations, Thank you. Could you also do a video on SQL Agent example with Langgraph?
Hey glad that you liked it :) will try to come up with a video on this...
super amazing! please do a lot more
Awesome presentation. You can make it more accessible by splitting the video into chapters.
Thanks for the tip!
Hi - I am like 15 years experienced now with plenty in data as well did a course on machine learning but this AI stuff am unable to structure and plan for. Can you suggest a roadmap i..e from where to start and move ahead and esp grab some certificates along the way.
You can start with Coursera ML and DL courses.
@@AritraSen Ok ...
@@AritraSen ML i akready know. Will DL lead to AI ? Please see if you can make a video of roadmp and how your videos can help travserse through it.
I have a playlist of DL with Pytorch here in this channel... You can start with it...
Also have a look at this post for roadmap - www.linkedin.com/posts/aritrasen_llm-rag-finetuning-activity-7162319532500480000-lIxW?
Thank you Aritra the series is really good.
Hey this series was super helpful can you recommend some sources to get some hands on this kind of use cases?
Follow any good tutorial series on object oriented programming in python in youtube or in web.
Thank you! Very clear explaination
Thank you for sharing such amazing topic.which is much required for realtime rag, i was stuck in such task from long time in my company task ,since i am starter in this sector, this will definitely help me build a better product ,where i work.kindly make some thing like , how to give a name to a rag system and other details as well.
More use cases will be so much appreciated man ❤️🙏👏
I did not understand this part def detach(states): return [state.detach() for state in states]
Hey, please have a look at this - pytorch.org/docs/stable/generated/torch.Tensor.detach.html
amazing tutorial , please upload few more tutorials
Amazing tutorial !
High quality content ❤👏 keep up the good work man!
I have a question. I have tried this example and also the example in the langgraph github. What I did was, I removed the tavily search tool and kept a RAG tool. So basically removed the "Researcher" tool. I also have another tool which just prints whatever the user wants. My RAG tool's context is to answer only about "LLM powered autonomous tools". I have set up a Vector DB and it's retriever. This is equivalent to your RAG tool answering questions about Japan and sports only. If you ask a question about a celebrity, it generates an answer about the celebrity, in the graph stream event, but the RAG tool actually answers with 'I don't know', which is what I want. How do I get the agent_node() to retrieve the answer from the RAG tool and not generate it's own answer?
hello thank u for the video! but you can integrate human in the loop on this graph structure to make a conversation
Yeah you can do that... Add a conditional edge whereever you want a user/human input... And decide the router edges based on user input.
Thank you!
Wonderful explanation! Can we add memory into the rag pipeline? is there any method to do it?
This is awesome :), please share the decks Bhatt
Hey, in the description of the video you have the notebook link... In the notebook required screenshots are given...
@@AritraSen Thank you
This is such a great video. Thank you! I was finally able to get more clarity on langgraph. You went through all the relevant lines of code needed to understand this graph pattern. Much appreciated. I request you to also explain the multi agent hierarchical agent teams as well, like how you did for supervisor pattern in this video. Also, if you could explain how to send back the final output of this graph to a vanilla FASTApi server (not via langserve), where the user query was passed to an API end-point. Does it need to be a streaming output? Can this even be done with plain FASTApi server? This would be helpful as well.
Thanks for the feedback!! I will try to look into the request definitely in near future.
Most useful video to begin generative AI thanks
I see you have very good knowledge of the subject but the whole point of making tutorials is to make it more intuitive to audience by grabbing attention all the time. I know that's hard but that's a constructive feedback hope you won't mind.
Thank you for this feedback, appreciated... Will try to make things more intuitive...
Woo ..this is very useful Aritra da 🎉
100th video. What a milestone. Cheers to 100 more. Continue the amazing work.
Only encouragement like this will keep me going :) thanks for it !!
Great work! Aritra, Could you please share the presentation you did as a PPT? That would be really useful as a notebook.
Hey sorry. That's not a regular ppt format which can be shared... You can take screenshots from the videos if you want...
This demo is awesome.
cloned repo but validation error on cell 2 relating to build_rag.py ValidationError: 1 validation error for HuggingFaceBgeEmbeddings model_name none is not an allowed value (type=type_error.none.not_allowed) can you give sample .env
Added dot env with the name env.sample... Once you clone please rename it to .env
@@AritraSen Thx! Those .env settings helped me get on the right 'path' to get it working :) FYI, I tried script with old gpt4 turbo preview and it didint work in my first run in that despite judging most docs as relevant, it transformed to a web query. When i switched to gpt-4o it was smooth as silk. Another FYI, i've got an old CPU and old gtx 1660ti GPU. When i first ran embeddings it took approx 9 minutes. Pip installed a gpu pytorch version and got to run on GPU doing embeddings in only 52 seconds!. Well worth changing the setting I guess even on old GPU. This was a great tutorial thank you. Have you looked at a RAPTOR RAG? Langchain did a good video on it here. th-cam.com/video/jbGchdTL7d0/w-d-xo.htmlsi=tMV2mFZHrphLFGEi I'm trying to get to just one complex RAG approach i can lock in as "good enough". Thanks again!
Pardon me for the sneezing , so annoying , kind of missed to edit the video :(
বড় জ্ঞানের সেশনে এমন ছোট ছোট ঘটনা ঘটতে থাকে।
:)
I appreciate this excellent breakdown Artira . The semantic chunker is something that could be a real breaktrough for technical documentations. One of the problems we keep facing is that "summarization" is often leaving out too many details (e.g.: techincal lists that are incomplete, instructions that should NOT be summarized are still summarized etc.).😶
Hey, glad you liked it :) One of the thing you can try out is giving instructions in the prompt and ask the model to think in a step by step fashion with those instructions.
Hello, I want to make a project like this: fast api or flask api + reactjs + (open source llm (Llama2, 3, mistral etc.). But I couldn't find any project that I can reference. Can you make such a project? I don't want to use openai. The project must be able to run without internet.
We already have a playlist - RAG LLM App with FastAPI and Gradio: th-cam.com/play/PLOrU905yPYXIqQLY6ulQqB8e414-DFuyd.html
@@AritraSen I think OP is looking to integrate a React front-end using an API. Gradio seems good for a demo but React is allot more customizable.
@@awakenwithoutcoffee hey sorry React is not my skill set...
@@AritraSen totally fine brother, keep on creating !
not finding your kaggle without the dataset cant do anything, what is your kaggle and the dataset you are using thanks for your videos are great you deserve more likes and subs
www.kaggle.com/code/aritrase/dlwithpytorch-seq2seq-machtrans-rnn-attention Here you go
can you share the paper? is attention is all you need?
It's a different one, here is the link - arxiv.org/abs/1409.0473
Nice tutorial. More langgraph examples would be appreciated
Added another one in the playlist :)
Excellent intro to langgraph, thx. "Conditional path" kwargs have changed since post
Glad you liked it, Is it on the conditional path? Can you please share the updated code for the viewers if possible...
@@AritraSen Was on definition of conditional edge cell: from langgraph.graph import StateGraph,END graph = StateGraph(AgentState) ### StateGraph with AgentState graph.add_node("agent", function_1) graph.add_node("RAG", function_2) graph.add_node("LLM", function_3) graph.set_entry_point("agent") # ###### HERE I CHANGED TO SOURCE, PATH AND PATHMAP ...conditional edges are controlled by our router graph.add_conditional_edges( source="agent", # where in graph to start path=router, # function to determine which node is called path_map={ "RAG Call": "RAG", "LLM Call": "LLM", } ) graph.add_edge("RAG", END) graph.add_edge("LLM", END) app = graph.compile()
@@IdPreferNot1 After This solution I got a different topic. -->Calling Agent Received question: Tell me about Japan's Industrial Growth Generated response: Topic='Array Data Type in JSON' -->Router--> Array Data Type in JSON -->calling LLM--> Every time I run code it gives Topic='Array Data Type in JSON' I tried every possible solution but could not solve it. Please Help me @AritraSen @IdPreferNot1
Very Good tutorial!
Thank you Aritra Video is really helpful
Glad it was helpful!
Nice one...Keep it up.
Hey Nice Video! I just wanted to ask that is your deep learning playlist enough for learning deep learning from scratch? Actually I have been scrambling through the internet for DL Courses and this one seems good.
Yes so far I have received good feedback... I start from very basic and then slowly move to complex, so that you don't feel overwhelmed :) only basic python knowledge is required.
But again no one would say my course is bad :) so I would recommend spend sometime and see how it's working for you... All the best :)
Dear Aritra, is this the same concept as Advanced RAG?
Yes you can say it, there are lot of advanced RAG concepts... This is one of them...
Very nice series of videos. I was reading a lot about these new concepts like HyDE but the way you have explained it so simply with this code snippet is wonderful. Thank you. Please keep up this good work. Looking forward to your next video...
🎉🎉🎉 greatgreat idea!
Nice explanation, but I am really waiting for some real-world examples. Maybe ahead in the series I might see some. As things become clearer with examples. And if you can include some more projects. It will be nice.
Go ahead with the series, there are real world examples also, starting with very basics to do the ground work... Slowly we will move to complex examples...
@@AritraSen Thanks Aritra.
Nice demonstration
Great video. I'm determined to complete all the videos in this playlist.
Awesome. You can do it :) my best wishes...