SD3_alphaGirls

This is a test version,

I adjusted the training code for diffusers, removed the clip and t5 occupation,

to train lora with a minimum of 12g video memory on the basis of 1024 resolution + 1024 instantiation

[28 AI portraits and half-body] This model trains in the state of bs3 with 23g video memory on the basis of 1024+1024, with random sampling to increase generalization.

Version: v1

*Parameters do not represent the optimal, for personal use only

Sampling Method(Sampler): euler normal

Enlargement Algorithm: 4x-UltraSharp |  4x_NMKD-Siax_200k  

Sampling Iteration Steps: 25-30

Redrawing Amplitude(Denoising): 0.1

Width 680, Height 1024

Prompt Word Relevance (CFG Scale): 4

I have placed the workflow of the raw image in the attachment,

*Please ensure the latest version of comfyUI, otherwise SD3's Lora will not take effect

======================================================

My communication group: QQ 852450267 Business consultation: QQ 95893026

======================================================