KZhead

Control Tone & Writing Style Of Your LLM Output

Greg Kamradt (Data Indy)
2023 ж. 17 Мам.
12 302 Рет қаралды
Жүктеу

Twitter: twitter.com/GregKamradt
Newsletter: mail.gregkamradt.com/signup
Code: github.com/gkamradt/langchain-tutorials/blob/main/data_generation/Instructing%20LLMs%20To%20Match%20Tone.ipynb
0:00 - Intro
0:49 - Main Tips
1:57 - Code Overview
2:50 - Method 1: Describe tone
4:21 - Method 2: Describe tone + examples
6:27 - Method 3: AI-Assisted
9:28 - Method 4: Technique Fusion
13:19 - Extra Credit: Loop through more people

Пікірлер

  • You're a legend mate. I learn so much in a few minutes of your videos. Thanks for sharing your valuable knowledge and helping shape the world.

    @top_1_percent@top_1_percent27 күн бұрын

    • Love it - thank you very much!

      @DataIndependent@DataIndependent27 күн бұрын

  • Cool, that reminds me of textual inversion in stable diffusion where they feed (more or less) random strings to the prompt and check how close the result is of the desired output.

    @lodepublishing@lodepublishing Жыл бұрын

  • Awesome job! Keep it up, you are definitely pushing out a ton of valuable and actionable material :)

    @giacomociarlini4557@giacomociarlini4557 Жыл бұрын

  • Shaan, levelsio and stephsmite really sounds like their human versions aside from hashtag and emojis. fascinating. Great work. Loving the breakdown of overview and then the code steps.

    @real23lions@real23lions Жыл бұрын

  • This is for sure one of the best videos I've seen this year! Amazing.

    @raonigabriel@raonigabriel Жыл бұрын

  • Awesome job! Cool! Great work

    @user-nf4rf8uu1s@user-nf4rf8uu1s5 ай бұрын

  • dude this content is insane...thanks for putting it out there!

    @ynginvestor89@ynginvestor898 ай бұрын

    • Nice! Awesome young - thanks for the support

      @DataIndependent@DataIndependent8 ай бұрын

  • insane output

    @eugeniocg3079@eugeniocg3079 Жыл бұрын

  • This is exactly what I needed, amazing insight!

    @aromagrande1914@aromagrande1914 Жыл бұрын

    • What’s the project you’re working on?

      @DataIndependent@DataIndependent Жыл бұрын

    • @@DataIndependentSome personality addition to a conversational LLM. I want to have a variety of personalities to prepare people for certain social interactions.

      @aromagrande1914@aromagrande1914 Жыл бұрын

  • Your tutorials are awesome!

    @tandooripaneertikka@tandooripaneertikka11 ай бұрын

    • Nice thank you

      @DataIndependent@DataIndependent11 ай бұрын

  • This content is god level. I found your channels thanks to the langchain cookbook course. So glad of that discovery.

    @brkajsmd5376@brkajsmd537611 ай бұрын

    • Nice!! Glad to hear it and thank you

      @DataIndependent@DataIndependent11 ай бұрын

  • You are amazing!

    @GeorgAubele@GeorgAubele25 күн бұрын

  • Thank you for sharing and explaining.

    @DreamsAPI@DreamsAPI11 ай бұрын

    • Thank you

      @DataIndependent@DataIndependent11 ай бұрын

  • Fascinating way of explaining and teaching , thanks for the amazing content By any chance , do you know how to add a prompt or a personality to my retrievalQa chain ?

    @haouasy@haouasy8 ай бұрын

  • 🔥🔥🔥🔥

    @eugeniocg3079@eugeniocg3079 Жыл бұрын

  • Lol. I trained the bot to write like me a couple of months ago and used some of this approach among others.

    @krisvq@krisvq Жыл бұрын

  • Would fine tuning yield better result or is that not guaranteed? Especially if you have large amounts of wirting examples

    @trackerprince6773@trackerprince677312 күн бұрын

  • hi awesome video could you make the next videos a bit louder? even on full vol on YT and my computer still sounds low, thanks!

    @logstriketech6360@logstriketech636011 ай бұрын

    • Thanks for the feedback, will do.

      @DataIndependent@DataIndependent11 ай бұрын

  • Why do you use the percent sign (%) for headers? Is this a best practice? Could you make a video about that or suggest a blog post?

    @borisrusev9474@borisrusev94742 ай бұрын

    • I picked it up from Greg Brockman here kzhead.info/sun/otmtk6usmaCDqIk/bejne.html tbh I haven't seen a performance increase but I need to test it out more

      @DataIndependent@DataIndependent2 ай бұрын

  • Gr8t stuff! It solved 90% of a project I am working on. Next step is to create a virtual persona with the COMBINED stiles of a given group of authors. For example: define the tone and style of an extreme right-wing political author, by combining many moderate to extreme right-wing authors. Then tweet away!!

    @cag6825@cag682510 ай бұрын

    • Nice! I don't love the controversy spreading on twitter, but I love the exercise in AI tone management! Have fun exploring

      @DataIndependent@DataIndependent10 ай бұрын

  • The "Control Tone and writig Style for LLMs " playbook is not n the GitHub repo..

    @kterry697@kterry69711 ай бұрын

  • What about fine tunning on the tweets dataset? Would It work?

    @isako44@isako44 Жыл бұрын

    • That would work but you would need a massive amount of tweets. At the beginning he said this is another option.

      @coolmcdude@coolmcdude Жыл бұрын

  • #interesting #generativeAI #LLM

    @techie_guy@techie_guy11 ай бұрын

  • What a wonderful explanation, when i trying to use the tweepy API I think the free version is not enough no more=(

    @rafacanseco@rafacanseco Жыл бұрын
KZhead