sd-scripts

sd-scripts copied to clipboard

SD3 different inference result using sd3_minimal_inference.py and diffusers

4

Using sd3_minimal_inference.py and diffusers to inference open source sd3 model can get totally different result, is it because of the scheduler? it seems that you use your own-written scheduler rather...

SatMa34

Update train_util.py to allow for using SGD without momentum

7

Using momentum increases VRAM requirement by quite a bit, since you have to keep track of past gradients. momentum=0 might not be optimal, but is nice to have if you're...

ddpasa

Hello! This is a quick fix that allows sample images to be generated for those (like me) who are on Mac or otherwise don't have CUDA. It just adds if...

millie-v

SD3 training error

6

hello i was trying the SD3 branch for testing sd3 training using https://github.com/bmaltais/kohya_ss/tree/sd3 this branch. it installed right and I got all the sd3 settings right. but when I start...

GoofyAi

support HunYuan DiT

19

# [WIP] This PR is a draft PR for contributors to check the progress and review the codes. ## This PR starts with a simple implementation by me for minimal...

KohakuBlueleaf

MultipleInvalid: extra keys not allowed @ data['datasets'][0]['subsets'][1]['is_reg']

6

For some reason I cannot use `is_reg` parameter for DreamBooth type training. I'm using the latest [commit](https://github.com/kohya-ss/sd-scripts/commit/0b3e4f7ab62b7c93e66972b7bd2774b8fe679792) from `dev` branch. My dataset config is the next: ``` [general] shuffle_caption =...

voplica-git

Adjustments to resuming training

3

Currently skips the resumed epoch if partway through and calculates global steps as only the number of steps into the current epoch. These changes make it resume mid epoch on...

Cauldrath

Does sdxl lora text encoder training, train both text encoders? or just one?

DarkViewAI

Given groups=1, weight of size [1536, 16, 2, 2], expected input[4, 4, 128, 96] to have 16 channels, but got 4 channels instead

1

Loading settings from /content/fine_tune/config/config_file.toml... /content/fine_tune/config/config_file You are using the default legacy behaviour of the . This is expected, and simply means that the `legacy` (previous) behavior will be used so...

hieusttruyen

[Feature request]: Add Kolors Support

3

It is SDXL architecture with a different noise scheduler(Kolors Scheduler?) and text encoder (ChatGLM3-base). Would be great to be able to fine tune it as it is better than any...

maybleMyers