sdxl lora training issue

[vesperialuna]

hi ben,

i have encountered this error on 2 separate pods when trying to train sdxl lora:

not sure what's going on. i have tried reducing the epoch from 50 to 25, i have also tried reducing source image from 50 to 30. any thoughts? thanks in advance.

[vesperialuna]

nevermind ben i think i solved my own issue lol. i cropped all the reference images to 1024 x 1024, rename the images to a uniform naming convention, and used a higher vram runpod (20gb vram) and it works now. sorry to bother you.