The model was trained on OneTrainer using two layers: single_transformer_blocks.7.proj_out and single_transformer_blocks.20.proj_out.

Lora rank16 was used, but the similarity was not 100%, suggesting that more training steps might have been necessary. Therefore, the Lora strength is recommended to be between 1.0 and 1.5+.

For the opt90 and opt95 versions, the kohya resize lora script was utilized.

reference: https://www.reddit.com/r/StableDiffusion/comments/1f523bd/good_flux_loras_can_be_less_than_45mb_128_dim/