Type | |
Stats | 61 476 |
Reviews | (18) |
Published | Apr 25, 2024 |
Base Model | |
Training | Steps: 900 Epochs: 20 |
Usage Tips | Clip Skip: 2 Strength: 0.7 |
Trigger Words | preyas, bakugan, humanoid, blue body, purple body, blue scales, red eyes, sharp teeth, tail, fins, claws, spikes |
Hash | AutoV2 4139DA8D9D |
Preyas from Bakugan Battle Brawlers. Sassy comic relief chameleon-fish-lizard-man. Again.
Pony XL version trained on Pony Diffusion V6 XL.
This was the last epoch from a training session cut short halfway. Luckily enough it works fine from the pics I generated.
Recommended weight of the LORA itself is 0.7.
Just like the 1.5 one, lol.
main tags:
preyas, bakugan, humanoid, blue body, purple body, blue scales, red eyes, sharp teeth, tail
optional tags:
fins, claws, spikes
P.S. be more mindful with the way you prompt with Pony XL models, they break in ways that was never possible in 1.5 models before. Anything from a Clip Skip that isn't 2, one too many weighted tags, or a few tags in the wrong place might end up generating broken pictures, unless that's your thing then congratulations.
{
"ss_output_name": "preyas",
"ss_sd_model_name": "stablediffusionapi/pony-diffusion-v6-xl",
"ss_network_module": "networks.lora",
"ss_optimizer": "prodigyopt.prodigy.Prodigy(decouple=True,weight_decay=0.05,betas=[0.9, 0.99],d_coef=1,use_bias_correction=True)",
"ss_lr_scheduler": "cosine",
"ss_clip_skip": "None",
"ss_network_dim": 8,
"ss_network_alpha": 4,
"ss_epoch": 20,
"ss_num_epochs": 42,
"ss_steps": 900,
"ss_max_train_steps": 1800,
"ss_learning_rate": 1,
"ss_text_encoder_lr": 1,
"ss_unet_lr": 1,
"ss_noise_offset": 0.0357,
"ss_adaptive_noise_scale": "None",
"ss_min_snr_gamma": 6,
"ss_training_started_at": "2024-04-22T14:40:09.939Z",
"ss_training_finished_at": "2024-04-22T17:02:10.137Z",
"training_time": "2h 22m 0s",
"sshs_model_hash": "8d2ca1cc42084f10172e7317d59c4a13120ed429b050f19329821942aa4c4041"
}