Heres two models. On i start create with 64/32 dim/alpha, but learning is but unfortunately, the training was interrupted, so I managed to download only 14 epochs out of 20. The second version has already been successfully trained (20\20). But in it I tried to use the 4/32 dim/alpha ratio. This is how this ratio was recommended in the civitai training system itself.

Files with SHA256: 3dfc1df109242128023742390b9ee244a7d13a5b80dad369dfead44f01ba7eac