@Dr. Furkan Gözükara does NOT-raining text encoder save memory and speed? Is it worth turning off TE
@Dr. Furkan Gözükara does NOT-raining text encoder save memory and speed? Is it worth turning off TE learning?
ROh just noticed someone asked the same question
Fit is turned off by default
Fit increases vram slighty and reduces training speed slightly
RAlso, where do you explain the block training --block_lr 1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5,1e-5
EI had Very good success using control net Depth, Canny and openpose to upscale and bring out details in an old photo from this
Fuse config from patreon
ROk good to know
EI wish there was a better Canny model for detecting edges
Ehad to mess with it a lot
Fye we need better sdxl controlnet
Ealso waiting for QR control net for sdxl, I cant seem to properly upscale 1.5 images from 512 to higher ones without losing details
SThank you so much!
S@Dr. Furkan Gözükara this is the new one that replace the real vision one? https://youtu.be/AY6DMBCIZ3A?feature=shared can I do it with runpod?
YouTubeSECourses
Updated for SDXL 1.0. How to install #Kohya SS GUI trainer and do #LoRA training with Stable Diffusion XL (#SDXL) this is the video you are looking for. I have shown how to install Kohya from scratch. The best parameters to do LoRA training with SDXL. How to use Kohya SDXL LoRAs with ComfyUI. How to do checkpoint comparison with SDXL LoRAs and m...
Fyou can do with runpod
Fi suggest you to use dreambooth config from patreon
Fand use that in dreambooth tab
Fbetter than lora
Syes I’m a supporter
Fye please use latest config from patreon
Fmuch better
Fafter loading config
Fjust set your folder paths
Fdmn kaggle not working with sdxl trying to solve
Sbeen a rough week for training lol
SI’m watching your tutorial right now 
Son this tutorial how is the procedure on runpod?
Snevermind I found the instructions on the github link
S
S33k steps later
Ffinally found the kaggle error
Fvery high quality
S@Dr. Furkan Gözükara more detail with 100+ hi res steps
S
S
Sthese are again with 0 class images
S220 professional photos, cropped but not down sized, trained at 768, xformers, 8bit adam, bf16, 150 epochs
Ldreambooth?
Syep
Lwhat do you by "cropped but not down sized,"
Sso i will crop them in closer to the subject so the face stands out more but i don't make them 1080x1080 or anything like that
Lwhat setting did you use for "Save Model Frequency (Epochs)"?
Si just put 50 but i liked the whole 150 best
Sbut so some of my photos i used were closer to 4000x5000 pixels
Lyour models looks amazin
S@Dr. Furkan Gözükara I'm having thiis error "torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 19.68 GiB total capacity; 12.33 GiB already allocated; 19.75 MiB free; 12.42 GiB reserved in total by PyTorc"
Si've been testing for fastest/ laziest way to make good models
SI already check "Gradient checkpointing"