Hey guys, does anyone know how to install tensorRT and adetailer on kaggle?
Hey guys, does anyone know how to install tensorRT and adetailer on kaggle?
T@Dr. Furkan Gözükara I've noticed that for your most recent tests you used Koya SS GUI for training, but on Patreon I've only been able to find setup scripts for the Dreambooth extension on RunPod. I don't see any reason I couldn't manually run Koya on a RunPod instance, but I'm hesitant because of the issues with Dreambooth only working with specific versions. Does Koya have similar issues? Also, what advantages does Koya have for training on individual people compared to the A1111 Dreambooth extension?
TSo far every model I've trained (using RealisticVision as a base) has given me outputs that are consistently blurry
Fi know adetailer
Fbut tensorRT wouldnt work
Fprobably
Fsince GPU there is old
Fkohya is seperate
Fso works in all cases
Fon patreon i have all
Fwhat you need? runpod kohya?
Fmust something wrong
Fyesterday i did a training on an anime sdxl model
Fresults are amazing
Ftesting
F
MGood day, can I run your SOTA Image Captioning Script on Kaggle?
MIs there a way to install it?
Fit depends which one you want to run?
Fyes you can install adetailer from extensions tab too
MWell, Idk which one is the best for captioning photos that feature a person.
TYup. I want to experiment with this for training but don't want to be limited by my local GPU. I did see one for Kaggle but I wasn't sure how limited their GPU selection is
Fi suggest runpod
Fkaggle is restrictibve
Fcan you pay few dollars?
Fkaggle is limited for that
Fyou need runpod
Mdo you mean it doesn't have enough VRAM?
MYou are right, if I want to use Vit-bigG-14 + blip2-2.7 I need 24GB VRAM
Fyes
Frunpod is good choice
Falso LLaVA you need 24 gb gpu
FLLaVA is very powerful too for caption
F
MOne more question: I'm also using your fooocus kaggle notebook, but it only uses 1 GPU.
Fno such script supports dual gpu atm
MIs this still starlightXL?
Fyes
Fi am also testing reg images affect right now
JHi guys how can I get good words (letters)? I’d like a message
MThe only one I use is https://civitai.com/models/176555?modelVersionId=214296
Introducing Harrlogos v2.0! Solving Stable Diffusion text generation one LoRA at a time! THIS DOES NOT REQUIRE CONTROL NET TO WORK Harrlogos now on...
Fcontrolnet best option
Fye this model also superior for text
JWhat should I upload on controlnet ?
JMm no idea how to use it
Fyou need to upload text but yes that would be hard
Fyou also need to upload pose
Fso combine both the main image + text
Fin controlnet
