You could use the local version of Stable Diffusion to do this with a checkpoint model trained on dreambooth for those who do not like using Colab, as you only need an Nvidia GPU with at least 10 GB of VRAM. I have already tested, most of the process, but cannot complete the final step of the training as only have a RTX 2080 at present.
Thanks for your video :) I have a question, Is there a way to connect the Dreambooth to the diffusionbee application on the Mac? If so, how do you do it? Since I saw you go to stablediffusion which is in your browser.
So far, no. The developer is working with all the interruptions that the holidays and being with family etc brings. You can, however, create your ckpt models in dreambooth and then import them into DiffusionBee to use in your renders, which is what I do. I literally have nearly 70 models that I switch out with in DiffusionBee (1.5.1) and that includes different iterations of trained dreambooth models where I test how good the training is.
Thank you for your comment. This is an interesting statement. The graphics cards run around $1000-$1800 or more. Then you have to have the system to put it on. I totally understand the desire for that, but for a lot of people, I think the free version of Google colab will work just fine. Def nice to have that speed on your own machine though.
Sometimes this happens when you don't have enough ram memory. If you are using the google colab for it, monitor this thorugh the Ram and Disk info in the upper right corner of your google colab. You can also click down from there to see "View resources" and this will show you when the memory spikes and if it hits your max level.
You could use the local version of Stable Diffusion to do this with a checkpoint model trained on dreambooth for those who do not like using Colab, as you only need an Nvidia GPU with at least 10 GB of VRAM. I have already tested, most of the process, but cannot complete the final step of the training as only have a RTX 2080 at present.
Love it. Thanks.
Thanks for your video :) I have a question, Is there a way to connect the Dreambooth to the diffusionbee application on the Mac? If so, how do you do it? Since I saw you go to stablediffusion which is in your browser.
So far, no. The developer is working with all the interruptions that the holidays and being with family etc brings. You can, however, create your ckpt models in dreambooth and then import them into DiffusionBee to use in your renders, which is what I do. I literally have nearly 70 models that I switch out with in DiffusionBee (1.5.1) and that includes different iterations of trained dreambooth models where I test how good the training is.
i would just buy a 3090 or 3080 to try to run it on my pc locally
Thank you for your comment. This is an interesting statement. The graphics cards run around $1000-$1800 or more. Then you have to have the system to put it on. I totally understand the desire for that, but for a lot of people, I think the free version of Google colab will work just fine. Def nice to have that speed on your own machine though.
I keep getting errors at the training part
Sometimes this happens when you don't have enough ram memory. If you are using the google colab for it, monitor this thorugh the Ram and Disk info in the upper right corner of your google colab. You can also click down from there to see "View resources" and this will show you when the memory spikes and if it hits your max level.