Follow me on X/Twitter for exclusive content: https://x.com/Citron_Legacy 😁
Please checkout all my Dragon Ball Loras:https://civarchive.com/user/CitronLegacy/models?tag=dragon%20ball
--- OLD Comments ---
My first success with an SDXL Lora. Please let me know what you think!
Also if you have any SDXL training advice or checkpoints you think I should use for image generation please let me know! I'd prefer a more anime and less realistic style if possible.
I don't have a strong enough GPU for SDXL so I have to use TensorArt's website to generate all the images so the generation data might be a bit messed up :(
I'll give 200 buzz to the first 10 people that post a picture on my Super Saiyan 4 Caulifla Lora. (Limit 200 buzz per account) All Ten posts are taken! Thank you everyone!
Also there is a 2000 Buzz image creation bounty for this Super Saiyan 4 Caulifla Lora here:
https://civarchive.com/bounties/589 - Bounty is over. Thanks to the participants and congrats to the winner!
Version 4 SUPER SAIYAN 4 update!
Please considering listening to Ka Ka Kachi Daze on Youtube when you use this for the first time.
Caulifla's SSJ4 supports a bunch of random changes. I gave her a black tube top and also a black jacket in some of these images.
I neglected to add tail to all my examples but it can easily be added/removed.
As prodigy Saiyan warrior she is able to achieve even more forms with just a little prompting!
You can change her hair color by simply specifying it.
Here are some example hair colors. Super Saiyan 5 lets goooo! lol
I'm running out of space to describe this model. I'll put the training data in the "About this version" section.
V3 Update (10/26/2023): V3 is trained with a new dataset that incudes interesting poses and different outfits. Overall it is more stable but the art style might be a little too strong?
Caulifla is easily my favorite character in Dragon Ball so I'll keep making updates if I can find good images.
I'll add in a Super Saiyan Caulifla once the base form is good.
Last few images are prototyping my Saiyan_Armor_DB Lora. I'll upload it once its finalized.
Weight: 0.7
Trigger: Caulifla_DB <-- Same as the Lora file name, isn't that easy to remember?
Extra prompts: bare shoulders, midriff,
Trained on 33 images
Description
[sdxl_arguments]
cache_text_encoder_outputs = false
no_half_vae = true
min_timestep = 0
max_timestep = 1000
shuffle_caption = true
lowram = true
[model_arguments]
pretrained_model_name_or_path = "stabilityai/stable-diffusion-xl-base-1.0"
vae = "/content/vae/sdxl_vae.safetensors"
[dataset_arguments]
debug_dataset = false
in_json = "/content/LoRA/meta_lat.json"
train_data_dir = "/content/drive/MyDrive/lora_training/datasets/Super_Saiyan_4_Caulifla_SDXL"
dataset_repeats = 10
keep_tokens = 1
resolution = "1024,1024"
color_aug = false
token_warmup_min = 1
token_warmup_step = 0
[training_arguments]
output_dir = "/content/drive/MyDrive/kohya-trainer/output/Super_Saiyan_4_Caulifla_SDXL"
output_name = "Super_Saiyan_4_Caulifla_SDXL"
save_precision = "fp16"
save_every_n_epochs = 1
train_batch_size = 4
max_token_length = 225
mem_eff_attn = false
sdpa = true
xformers = false
max_train_epochs = 10
max_data_loader_n_workers = 8
persistent_data_loader_workers = true
gradient_checkpointing = true
gradient_accumulation_steps = 1
mixed_precision = "fp16"
[logging_arguments]
log_with = "tensorboard"
logging_dir = "/content/LoRA/logs"
log_prefix = "Super_Saiyan_4_Caulifla_SDXL"
[sample_prompt_arguments]
sample_sampler = "euler_a"
[saving_arguments]
save_model_as = "safetensors"
[optimizer_arguments]
optimizer_type = "AdaFactor"
learning_rate = 0.0001
max_grad_norm = 0
optimizer_args = [ "scale_parameter=False", "relative_step=False", "warmup_init=False",]
lr_scheduler = "constant_with_warmup"
lr_warmup_steps = 100
[additional_network_arguments]
no_metadata = false
network_module = "networks.lora"
network_dim = 64
network_alpha = 32
network_args = []
network_train_unet_only = true
FAQ
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.