100% Private & Local PDF ChatBot (without langchain)

Поделиться
HTML-код
  • Опубликовано: 28 дек 2024

Комментарии • 52

  • @abhishekkrthakur
    @abhishekkrthakur  Год назад +8

    Please subscribe and like the video to help me keep motivated to make awesome videos like this one. :)

  • @PRIYAKUMARI-xp2qr
    @PRIYAKUMARI-xp2qr Год назад +20

    What a genius statement: Open source models are going to beat the closed source models soon, can't wait.💫

    • @prafullsharma8052
      @prafullsharma8052 Год назад +1

      in dreams

    • @abhishekkrthakur
      @abhishekkrthakur  5 месяцев назад

      @@prafullsharma8052 still in dreams? :)

    • @prafullsharma8029
      @prafullsharma8029 5 месяцев назад

      @@abhishekkrthakurreality is they won’t beat closed source if they continue using closed source for data tagging . You and I both know , since you are with HF, you won’t admit it .. may be focus on doing something better instead of replying me

    • @abhishekkrthakur
      @abhishekkrthakur  5 месяцев назад

      @@prafullsharma8029done with better things. thanks for the suggestion. now its time to do the useless thing and respond to your useless comment. i believe you didnt read the paper where you can read what kind of data was it trained on. anyways, do something useful, its been a year and you are stuck on this without any proper references 🤣🤣🤣

    • @prafullsharma8029
      @prafullsharma8029 5 месяцев назад

      @@abhishekkrthakur right , I believe you haven’t read the same, most of these models are just boasting with useless Benchmarks…. Using benchmarks according their model’s suitability.

  • @shivamroy1775
    @shivamroy1775 Год назад +7

    Amazing content, love it when the videos are long and Abhishek takes the time to explain and code everything.

  • @byob801
    @byob801 Год назад

    Great video, I liked seeing you put this together and explain along the way. Very interesting to a novice like me. Thanks so much!

  • @pasqualmas8584
    @pasqualmas8584 Год назад

    Crazy video! Subscribed!
    If it could be i would like that in the first minute you show the result for knowing better what will you do in 38minutes 😂

  • @dr.aravindacvnmamit3770
    @dr.aravindacvnmamit3770 11 месяцев назад

    Very clean explanation!!!!

  • @abirmohammed17
    @abirmohammed17 Год назад

    @abhishek Thank you for this tutorial. Can you please suggest what should be added in the parser if the pdf file is know to have tables with data only?

  • @krzysztofwos1856
    @krzysztofwos1856 Год назад

    The `sentences.append(window)` in the `embed` function should go before the `if`.

  • @Sudip_Sarkar_Charles_Edwards
    @Sudip_Sarkar_Charles_Edwards Год назад

    Wow. Keep it up.

  • @quantadotonium3654
    @quantadotonium3654 9 месяцев назад

    Amazing! Thank you.

  • @yusufkemaldemir9393
    @yusufkemaldemir9393 Год назад +1

    I am confused with the fine tuning llama 2 video and %100 pdf video you have. I have hundreds pages of pdf that I want to make query and get correct answer. Do I need to follow your fine tuning llama2 video for peft and then follow this pdf video? Or just use this pdf video as reference? What is your opinion?

  • @VipulChauhan-s4u
    @VipulChauhan-s4u Год назад

    can you explain in the line of if he pdf document contains a mixture of tables and text in sequence such that the tables and text paragraphs make sense when are embedded together?

  • @ikjb8561
    @ikjb8561 Год назад

    Abhishek, great videos! I know you are using rtx 3090 for your local setup. How are you able to load Falcon40B on it?

  • @DB-in2mr
    @DB-in2mr Год назад +1

    open killed the closed star 😎

  • @nikhilthapa9300
    @nikhilthapa9300 Год назад

    Thanks for the tutorials, could you please do tutorial on multiple instance learning using Neural networks ?

  • @vaibhavsaxena6482
    @vaibhavsaxena6482 Год назад

    Hidden gem. Can we also use this for other languages?

  • @dreamhunter999
    @dreamhunter999 Год назад

    So cool!👍

  • @aminkarbassi
    @aminkarbassi Год назад

    Thanks for this video. You say that you have the 7B model running on your local machine, can you please specify the specifications of your local machine?

  • @iknoorsingh7454
    @iknoorsingh7454 Год назад

    Great work, Abhishek! Any idea how to prevent randomly generated output by 7B model - Is prompt engineering the only solution?

    • @sauravchat2002
      @sauravchat2002 Год назад

      Even I have the same question. Wasted a whole day trying to figure this out but without any luck.

    • @themusicalmarvel5149
      @themusicalmarvel5149 Год назад

      Retreival Augmented Generation. The generated outputs should be checked against a vector DB to prevent hallucinations.

  • @chiragv2294
    @chiragv2294 Год назад

    Sir, What is cheaper option? Using open ai API or deplying falcon 40 b on cloud?

  • @muhtalhakhan
    @muhtalhakhan 9 месяцев назад

    how to get this kind of detailed terminal?

  • @nullvoid12
    @nullvoid12 Год назад

    Want to work for hunggingface, how can I apply and get noticed? Awesome content btw!

  • @alexdelaiglesia1926
    @alexdelaiglesia1926 Год назад +1

    I couldn't ask this on time during the stream xD. Could you share your thoughts about the debate of using proprietary APIs vs deploying your own open source LLM (I mean, taking into account costs, privacy, etc.)? Thanks!

    • @nayanjitsarkar9281
      @nayanjitsarkar9281 Год назад +1

      I was about to ask the same question. Would really love to know from you Abhishek sir

    • @yazidridwan6917
      @yazidridwan6917 Год назад

      this !, i'm considering making my thesis on this

  • @sauravchat2002
    @sauravchat2002 Год назад

    Quite curious to know your local machine configuration. Is that something that can be disclosed? Or have you already mentioned it in the previous videos? The reason I am asking is currently I am operating with a 8 GB ram on windows 10 and that feels like config from a different era :) on which most of the things won't run.

    • @abhishekkrthakur
      @abhishekkrthakur  Год назад +1

      i run code for my videos on a machine with 32gb ram and 3090 gpu.

  • @mohsenghafari7652
    @mohsenghafari7652 10 месяцев назад

    hi. please help me. how to create custom model from many pdfs in Persian language? tank you.

  • @shivamanand4334
    @shivamanand4334 Год назад

    Can we replace my gpt 4 calls in my deployed flask app with this? Offering clients far more usage for same price?

  • @mariocuezzo8027
    @mariocuezzo8027 Год назад

    how i could add a webui to this code?

  • @ryanpopa3312
    @ryanpopa3312 Год назад

    do you store the code in github?

  • @pratiksaria5830
    @pratiksaria5830 Год назад

    Thank you very much sir

  • @musifmuzammir354
    @musifmuzammir354 Год назад

    Can't we reduce the hallucination by reducing the temperature value??

    • @abhishekkrthakur
      @abhishekkrthakur  Год назад

      sure. playing around with parameters is upto the viewers and is almost always helpful :)

  • @joeljunior7426
    @joeljunior7426 Год назад

    thank you master.

  • @zerophase2338
    @zerophase2338 Год назад

    Nice flexibility on offer here.

  • @soodisin
    @soodisin Год назад

    Can you please post the code as well

    • @abhishekkrthakur
      @abhishekkrthakur  Год назад +1

      updated in description :)

    • @soodisin
      @soodisin Год назад

      @@abhishekkrthakur What are your views on Azure Open AI vs Open AI. My knowledge is Azure uses Open AI models as well, so is there a chance that the data on Azure Open AI has the same privacy issues as Open AI?

    • @MuhammadGhazalli
      @MuhammadGhazalli Год назад

      @@soodisin Let me help to answer, for Azure they will capsulate your data and won't gather it. Your data will not move anywhere.

  • @yusufkemaldemir9393
    @yusufkemaldemir9393 Год назад

    @abhishekkrthakur this is not running on M2 Macbook locally. I am not using Docker.

  • @vobbilisettyveera2973
    @vobbilisettyveera2973 Год назад

    raise ConnectionError(e, request=request)
    requests.exceptions.ConnectionError: HTTPConnectionPool(host='0.0.0.0', port=6000): Max retries exceeded with url: / (Caused by NewConnectionError(': Failed to establish a new connection: [Errno 111] Connection refused'))
    please help me with this