Please checkout my other Shokugeki no Soma Loras: <a target="_blank" rel="ugc" href="https://civarchive.com/collections/110377">https://civarchive.com/collections/110377</a> <hr />Outfits<ol><li>Chef - <code>chef, chef hat</code></li><li>Witch - <code>purple cloak, skull accessory</code></li></ol><hr />I confirmed a new technique for Lora training works with this! Sadly Sadatsuka does not have much screentime or fan art so there are only about 24 images of her. I duplicated 16 of the best images bringing the working dataset to 40 images. This gives more emphasis on these particular images and also helps reach a good dataset size. <hr />Trained on: 24 imagesTraining Model: Animefull-final-pruned-fp16flip_aug: FalseNum of Repeats: 20 (and 16 duplicated images)Unit is Epochs or Steps: EpochsNumber of Epochs or Steps: 10Training Batch Size: 2Network Dim: Dim=32; Alpha=16Lora Creation Process took: 0:38:37

Sadatsuka Nao (Shokugeki no Soma)

Sadatsuka_Nao_Shokugeki.safetensors

v1.0

Please checkout my other Shokugeki no Soma Loras: <a target="_blank" rel="ugc" href="https://civitai.com/collections/110377">https://civitai.com/collections/110377</a> <hr />Outfits<ol><li>Chef - <code>chef, chef hat</code></li><li>Witch - <code>purple cloak, skull accessory</code></li></ol><hr />I confirmed a new technique for Lora training works with this! Sadly Sadatsuka does not have much screentime or fan art so there are only about 24 images of her. I duplicated 16 of the best images bringing the working dataset to 40 images. This gives more emphasis on these particular images and also helps reach a good dataset size. <hr />Trained on: 24 imagesTraining Model: Animefull-final-pruned-fp16flip_aug: FalseNum of Repeats: 20 (and 16 duplicated images)Unit is Epochs or Steps: EpochsNumber of Epochs or Steps: 10Training Batch Size: 2Network Dim: Dim=32; Alpha=16Lora Creation Process took: 0:38:37

Files with SHA256: 130461b89af6d1495a1446fe01699636646d7d825385c3491a77dff0a2de22e6