This model was made after I saw https://civitai.com/models/943256/johannes-frederik-engelbert-ten-klooster-style, which looks great but is quite large, so I cannot use it. Therefore, I decided to train one myself, and this is the result.
It turned out to be quite different, with its own distinct style, and is smaller.
Most of the gallery prompt are stolen from the other Klooster model so that one can compare the output from the two models 😅🙏
More information about the artist: https://nl.wikipedia.org/wiki/Johan_ten_Klooster
LoRAs were trained on tensor, and you can find all the epochs here:
tensor. art/models/796080988972273897/Klooster-Var2-5Cos-D8A4-R20E10-2024-11-14-18:47:24-Ep-10
tensor. art/models/795855881146332795/Klooster-Var1-5Cos-D8A4-R15E10-2024-11-14-04:33:54-Ep-10
These are the settings I used for training:
For the one with border (13 public domain images): https://civitai.com/images/40389929
FLUX.1 - dev-fp8
Trigger: k1ooster
Repeat: 15 Epoch: 10 total Steps = (* 15 10 13)=1950
Unet LR: 0.0005 Scheduler: cosine Optimizer: AdamW
Network Dim: 8 Alpha: 4
Without border (11 public domain images): https://civitai.com/images/40389951
FLUX.1 - dev-fp8
Trigger: k1ooster
Repeat: 20 Epoch: 10 Total steps: (* 20 10 11)=2200
Unet LR: 0.0005
LR Scheduler: cosine Optimizer: AdamW
Network Dim: 8 Alpha: 4
The training set is slightly different and I removed all the borders. Also each epoch was trained for 20 instead of 15 steps.
1. Os direitos dos modelos repostados pertencem aos criadores originais.
2. Criadores originais que desejam reivindicar seus modelos devem contatar a equipe do SeaArt AI pelos canais oficiais. Clique para reivindicar
