Creating Embeddings and Concept Models with Invoke Training - Textual Inversion & LoRAs

Поделиться
HTML-код
  • Опубликовано: 15 ноя 2024

Комментарии • 34

  • @Javan_lima
    @Javan_lima 7 месяцев назад +3

    I was waiting for that 😂 thanks !!

  • @jeeperchris
    @jeeperchris 7 месяцев назад +4

    This was great! I would love to see a Lora setup of a specific object to help steer away from biases. For example a very specific species of a parrot (not a macaw) is almost impossible to pull out.

  • @keith_carmona
    @keith_carmona 7 месяцев назад +3

    Thank you. This was very informative.

  • @angelotsk3173
    @angelotsk3173 7 месяцев назад +2

    After watching this video yesterday, i created my first sdxl lora ... very nice results .. thanks for this great tutorial and thanks for developpers who made this great repo, easy to use and very intuitif

  • @SonnyNguyen
    @SonnyNguyen 7 месяцев назад +2

    Thank you! This is exactly what I was waiting for!

  • @MarcusLlewellyn
    @MarcusLlewellyn 7 месяцев назад +5

    Thanks for this! I've already experimented with training a lora, and like it as a starting point. A few papercuts here and there though.
    - Being able to browse for models and dataset files or paths from the UI would be nice! (Freakin' quotes! Why does Windows do this to us?)
    - Auto-captioning via LLaVA or similar could really accelerate dataset creation.
    - Is there some way stop and then resume training? Although I have a relatively beefy GPU, I can't always dedicate it to training for extended periods.
    - Documentation! But I think you know this, hence the video. 😄

    • @invokeai
      @invokeai  7 месяцев назад +2

      We agree with a lot of these! Contributions welcome to help us accelerate those :)

    • @flashpowr
      @flashpowr 7 месяцев назад

      put the pc to sleep it will pause it I found out.

  • @TimothyMusson
    @TimothyMusson 7 месяцев назад +2

    Thanks for all the great work that goes into Invoke, including the documentation and these excellent tutorials! :)

  • @unveil7762
    @unveil7762 7 месяцев назад +1

    please do more tutorial like this !! this was really useful !!

  • @LouisGedo
    @LouisGedo 7 месяцев назад +4

    👋

  • @MyAI6659
    @MyAI6659 7 месяцев назад +1

    Awesome. Love you guys. ❤

  • @mr.N00003
    @mr.N00003 6 месяцев назад

    we need more videos is this playlist sir.

  • @IvanoColtellacci
    @IvanoColtellacci 7 месяцев назад

    Great explanation. Will be this tool also available on Mac M1 machines ?

  • @brianwade2758
    @brianwade2758 4 месяца назад

    Want to train a lora for old-timey moustache styles. I'm sure you can easily imagine what I get when I prompt for a pencil-thin, toothbrush, or handlebar moustache. As interesting and comical as the images have turned out, I'd really like to be able to get a handlebar moustache on a guy without a handlebar sprouting from his face.

  • @JoergFrochte
    @JoergFrochte 6 месяцев назад

    Very helpful

  • @barthays5947
    @barthays5947 7 месяцев назад +2

    Thank you! We do interior exhibition design, we have a sketch style and we would like to take "technical drawings" in our 3D Design app and give them our Sketch style without a ton of work in PS. I think an embedding would get us there. Would love some help optimizing for my RTX 3070 8Gb

    • @barthays5947
      @barthays5947 7 месяцев назад

      I'm going through the process, I get an error in Invoke when I try to import the model: " Could not determine base type"

  • @Destroyes08
    @Destroyes08 7 месяцев назад +2

    Hello, I installed the system training and tried to launch it via the GUI but it tells me "command not found" I tried to figure out without any sucess, is there something I'm forgetting ? It seems the command doesn't exist in my system.

    • @invokeai
      @invokeai  7 месяцев назад

      You need to activate the venv first :)

    • @Destroyes08
      @Destroyes08 7 месяцев назад

      @@invokeai I did but I still have the same problem

  • @angelotsk3173
    @angelotsk3173 7 месяцев назад +1

    Nice Nice .. Nice

  • @AndyHTu
    @AndyHTu 7 месяцев назад

    I'd love to be able how to train my own LoRa for Styles, and Characters.
    I've been trying to learn how to write captions and learned about Structured Data for writing the captions where it needs to be organized and be consistent across every images and only add variables when it becomes important to train the models. That's something I'd love to get deep into. :)
    And of course what all those settings mean like Epochs etc. And when to use it.

  • @muhammadahsan572
    @muhammadahsan572 4 месяца назад

    hi, i am trying to train textual inversion SDXL on solana coin images. i have tried multiple runs but it is not learining anything. can anybody guide me what i can be doing wrong?? i followed the same steps as in the video

  • @SonnyNguyen
    @SonnyNguyen 7 месяцев назад +1

    So, is there a way to create a .bat file that help run the training GUI automatically instead of using console with commands to start "# Activate the new virtual environment: Windows: .\invoketraining\Scripts\activate " and "# From the invoke-training directory: invoke-train-ui"?

    • @invokeai
      @invokeai  7 месяцев назад

      Yes - we’d just need some contributions to help with installation and running more easily!

    • @angelotsk3173
      @angelotsk3173 7 месяцев назад +6

      @echo off
      set "venv_path=invoketraining\Scripts\activate.bat"
      call %venv_path%
      invoke-train-ui
      python --version
      pause
      -------
      Save this as a bat file and put it in your folder , D:\invoke-training

    • @angelotsk3173
      @angelotsk3173 7 месяцев назад

      And this to launch the Dashboard:
      @echo off
      cmd /k "M: & cd M:\invoke-training\invoketraining\Scripts & .\activate & cd.. & cd.. & tensorboard --logdir output & set url="localhost:6006/" & start chrome %url%
      pause
      ------
      change M: with your Hard-disk where invoketraining is located, change also the path .. Save it as a bat file

    • @greatjensen
      @greatjensen 7 месяцев назад

      @@angelotsk3173 All I get when executing the new .bat is:
      'invoke-train-ui' is not recognized as an internal or external command,
      operable program or batch file.
      Python 3.10.10
      Press any key to continue . . .

    • @SonnyNguyen
      @SonnyNguyen 7 месяцев назад

      @@angelotsk3173 wow that works! Thank you so much!
      @invokeai This is it!

  • @jasonstapleton4091
    @jasonstapleton4091 3 месяца назад

    i have not used anything ai in a while and am starting from scratch. i want to train several faces of myself and family to add multiple people to an images. Im not sure how to go about this. Any advice would be welcomed.

  • @boricuapabaiartist
    @boricuapabaiartist 7 месяцев назад

    Fantastic, did you make yourself an embedding for a cat on a fire hydrant,? 😂. And for that data set did you use canvas to inpaint the cats on top of the hydrants?