Hi thanks for reaching out, however the application will show the dataframe only but the changes are supposed to be in the dataframe, like if the date format is changed, also you can check the CSV generated in the backend.
Hey, great video, thanks to you I got it working. Is there a possibility to change the code, so that not the first modelfile is downloaded? I want to download Q4 Mistral for examble but the code gives Q2. I am pretty new to this, sorry if this is a silly question.
Thank you for reaching out !! Good question it is. In another video I have shared different ways to load the models, with Llama_cpp you can manually download the model of your choice and load that for inference.
Thank you sir , have you used open source LLMS such as Claude 3 , Llama 3 or Mistral ? what do you think of a fine tuned model ? some companies don't like closed source models (security ...)
Thank you for commenting ! I have tried to do it with Llama 2 and Mistral , well with 7B models, the response was not so good. However, I think with the higher parameter model, it must be possible.
Thanks Joy. Once the vector DB is loaded with schema and if we have some sample questions and corresponding SQLs how can we add it to the vector DB for better results ? Also does the vector DB keeps updating itself with more interaction or how can we improve by manually adding more questions and SQLs?
Thanks for reaching out ! You can update the metadata of the records in vectordb. I have another video on Vector DB, that might be of help. Vector DBs don't have any intelligence to update itself. But we can build logics around it to handle it
Thanks for reaching out!! It does not work for image based pdf, for that chatGPT vision is required. And have to pass the pages of the pdf as image in the prompt. I am making a video on that will share soon. Request you to stay connected!!
Thank you, Joy. I am a new learner on RAG. Would you mind show us a demo on using module with OpenAI (not AzureChatOpenAI) and use MySQL (with full database tables provided)? I hope can follow with you to build up the chatbot step by step successfully. Thanks!
Hey Joy, Can you plz help how can we use input as book cover image and find similar book search with the same code , also provide the code for the same , dataset??
Hi Neha, thanks for commenting, but this approach will definitely not work for your requirement, this implementation is for textual inputs, you would need a multi modal LLM to interpret the image input. I will try to cover that in one of my future videos.
Thank you so much, this explanation is great! This really help me a lot but, i'm stuck at adding my own gguf models to my project. Like when i'm trying to add it my code didn't detect it and downloaded the other version of the model id. Can i download the models manually from Hugging Face than downloading it from the script? Because the file i downloaded from the script, is not even a gguf file or any type of that.
Yes, that is absolutely possible, you can manually download and provide to the local file path in LlamaCpp, for CTransformers also the same is possible
Hello, thank you so much for this video. i have a question related of summarize questions in LLM documents. For example in vector database with thousands documents with date property, and i want ask the model how much document i received in the last week?
@@joymaitra5414 , thank you for the beedback, but the questions are illimited, the user can ask direct questions like "what is the latest document i have?" or he can ask "how many documents i didn't read ?", i don't know how can convert this question to vectorDB query.
Hey good video. I am trying to recreate this, created the .env all of that is situated, however when I run ui.py using streamlit, getting error AttributeError: module 'openai' has no attribute 'error'. Can you help??
I have two queries. It would be helpful if you can assist in that. 1. Template which you are using for prompt , in the first line you have provided f string. But however in the next lines you are not giving where you have {context}, {question}. So wouldn't that make it a string rather taking context and questions as variable? Or the model will understand that? 2. My second question is that I have tested the code. It is working for most part except Chat history. Even in the same session, it is not able to remember the previous questions. It is getting updated in the chat history though. But when u explicitly ask what is my previous question or ask like explain more on previous question, it is not taking the chat history rather giving halucinated questions or says it doesnt know previous question. Kindly clarify on the above two questions
Hi, Thanks for sharing the queries, for the 1st query, you are right, need to fix that. But for the 2nd question, this demo I have shown only the chatbot, while in the other view I have demonstrated how to utilize the memory in llm. Let me know, if you have further queries. ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE--TtE7otMHCc.html
Well I find the content spread all across the internet, trying to bring them together in one place. You can share your particular queries in my channel community, I will try to share my thoughts there.
thx the video! sorry my question but im a beginner in this code. What files I see github I tried to run on Codespace with no success. Is there any other file, enviroment info I should have to try and run your RAG app? Could you pls help me regarding this as a beginner?
hey this is really useful.can you explain how to extract text from pdfs for indic langugaes.these like to extract hindi languages we are not able to extract with regular pypdf or ocr techniques.kindly help with some codes.
Great one! Can you make a tutorial on, how can we finetune our custom dataset locally and use that finetuned model for getting domain-specific results locally?
great video Joy, do u think will be possible to have the model downloaded into a file and have a docker container that runs the webapp and interact using the model?
@joymaitra5414 Even with some API, could be accessible from Linux CLI and thst also be the endpoint to a web Interface as well. get the best of both worlds CLI and WWW. will be happy to work with u in this.
Hi Joy, thank you for this video. What is the advantage of using ctransformers library versus other github libraries available such as OpenLLM? Is it just a matter of personal preference?
Hi, thanks for your quey, more than preference i would say understanding the core, openLLM is like a wrapper created using the basics that tried to capture.
@@joymaitra5414 I prefer simple as well. I will try with just ctransformers to see if that improves implementation. Right now, too many wrappers out there haha.
