|
cardass9-300/600/900 ckpt = photo of cardass9 person with black hair |
|
|
|
c-ds9350 = photo of ds9cardassian person with black hair |
|
|
|
c-ds9500ckpt = photo of c-ds9 person with black hair |
|
|
|
cds9300/325/350 = photo of cds9 person |
|
|
|
cardass1an ckpt = photo of cardass1an person |
|
|
|
|
|
|
|
|
|
--- |
|
cds9(steps of training) |
|
cardass1an is the trigger word unless cds9 model which is cds9 |
|
or c-ds9 person for the class images included model |
|
OR ds9cardassian person with black hair for 350steps with classification |
|
model trained on 3 instance images of cardassians from ds9, 1000 steps no prior preservation no class images |
|
|
|
--pretrained_model_name_or_path=$MODEL_NAME \ |
|
--pretrained_vae_name_or_path="stabilityai/sd-vae-ft-mse" \ |
|
--output_dir=$OUTPUT_DIR \ |
|
--revision="fp16" \ |
|
--seed=1337 \ |
|
--resolution=512 \ |
|
--train_batch_size=1 \ |
|
--train_text_encoder \ |
|
--mixed_precision="fp16" \ |
|
--use_8bit_adam \ |
|
--gradient_accumulation_steps=1 --gradient_checkpointing \ |
|
--learning_rate=1e-6 \ |
|
--lr_scheduler="constant" \ |
|
--lr_warmup_steps=0 \ |
|
--num_class_images=0 \ |
|
--sample_batch_size=1 \ |
|
--max_train_steps=1000 \ |
|
--save_interval=500 \ |
|
--save_sample_prompt="photo of cardass1an person" \ |
|
--concepts_list="concepts_list.json" |
|
|
|
|