[Day 83] Summary of my PDF RAG from scratch

 Hello :)
Today is Day 83!


A quick summary of today:
  • Played around with llama-index and parsing, and then organised the github repo to be a bit more presentable


Below I will go over the project.


Used pdf: Lloyds Banking Group's Relationship T&C from their website, includes high-level information about accounts, and information about the user and the Bank's rights.

Embedding model: mixedbread-ai/mxbai-embed-large-v1 from huggingface

Language model: google/gemma-2b-it from hugging face

Repo structure:


PDF parsing: 

  1. For the main RAG chat (rag.py) ->  using PyMuPDF and langchain's RecursiveCharacterSplitter
  2. For the dev RAG (dev_rag.ipynb) -> llama-index and langchain's RecursiveCharacterSplitter

RAG chat

I call the main 'RAG chat' - these are preprocess_pdf.py and rag.py, and with them, using the terminal, a user can read, preprocess a PDF, 

Here is a demo I recorder from my terminal



Dev PDF chat:
I experimented a bit with chunk_size and chunk_overlap - the params of langchain's text splitter, and finally I settled on 1000 and 0, and those seemed to provide decent results when using llama-index as the PDF reader. Below are the sample queries and the answers provided by the model
However the results are not consistent, if a model starts to predict the words 'the context' the end would be that the context does not provide information about the query, but sometimes when I re-run the code, the answer actually gets generated. This could be due to several issues, but one idea that came to mind was adding beam search, because even if the model starts to generate 'the context ...' in one output, I believe in one of the others - the actual answer would appear. Actually, if we add return_context=True to the ask function, the top 1-2 context has the answer, but the model does not see it. Some other reasons - the way the context is provided (i.e. it is not clear to the model), bad query, bad base_prompt (how I tell the model to read for context), probably other reasons of which I am not aware at the moment (or I am not recalling them as reasons for problem).

Instructions for setup:
I see in other repos and just in general, setup instructions are added. I added as well and it looks nice haha

Closing thoughts:
This particular one uses out-of-the-box embedding and LLM, but the world out there might require me to make those from scratch too. I am still learning about RAG, I learned a lot about retrieval models from the ACL talk (Day 74, 75, 76), and I am looking forward to applying and seeing that knowledge in practice. 

The github repo link is at the top, any help is appreciated ^^

That is all for today!

See you tomorrow :)

Popular posts from this blog

[Day 198] Transactions Data Streaming Pipeline Porject [v1 completed]

Image
 Hello :) Today is Day 198! A quick summary of today: data streaming pipeline project [v1 done] Here is a link to the project's repo. Well ... I did not know I can do it in a day (~14 hours) after yesterday's issues but here we are. Turns out in order to insert the full (~70 variables with nested/list structure), I need the proper pyspark schema. And yesterday I did not have that and that is why when I was reading data in the kafka producer I was getting NULL in the columns - my schema was wrong. Well today I not only fixed the schema for the 4 variables I had yesterday, but included *all* the variables that come from the Stripe API ~ 70 (for completeness).  When I run docker-compose, the data streams and is input into the postgres db (and is still running). Unfortunately, the free Stripe API for creating realistic transactions has a limit of 25, so every 3 seconds, 25 new transactions are sent to the db. It has been running half the day (since I got that set up) and as I am wr
Read more

[Day 107] Transforming natural language to charts

Image
 Hello :) Today is Day 107! A quick summary of today: created text2chart - transforming natural language to charts [ github ] [ webapp ] I am super sick, but in the around 3 hours of my fever not being high, I managed to study a bit. I was again thinking about creating some nlp project but was stuck on the idea of how to host a model. After looking around I found this article where they create a similar app called chat2vis  and they use codellama. I know codellama is an open source model that I can get from huggingface and I wondered, how did they host it. Turns out - they ask the user to input a huggingface api key, which then allows for querying and generating content. (AWESOME! this is applicable to some of the other PDF chat apps that I have, but I will look to make an app of them later).  My fever eased a bit, and I sat on my chair to do some work ^^ Here is a breakdown of the app. I used streamlit. Also, 2 default datasets during dev: Financial Statements of Major Companies(2009-
Read more

[Day 54] I became a backprop ninja! (woohoo)

Image
 Hello :) Today is Day 54! A quick summary of today: Read through cs231n's neural network lectures [1] [2] [3] Spent 8 hours going over Andrej Karpathy's backprop ninja tutorial After yesterday's confidence boost in doing backprop, I came in today with the intention of becoming of, as Andrej Karpathy calls it, a backprop ninja.  The cs231n NN lectures were the usual about how NN became to be, param initialization, activation functions, loss functions, regularization and optimizers. Nothing out of the ordinary, but still was good to reassure myself of that knowledge. But that was definitely not the main even of today. As I mentioned, I wanted to become (or at least attempt to become) a backprop ninja today. I have watched that particular tutorial twice, the 1st time I had no idea what is going on, the 2nd time I could follow, and now it was the 3rd time - and now, I wanted to participate and actually see the why and the how of the things we have to do, when we do manual back
Read more