Use mina ashido on your prompt, this was trained on this collab: <a target="_blank" rel="ugc" href="https://colab.research.google.com/drive/1bFX0pZczeApeFadrz1AdOb5TDdet2U0Z?usp=sharing">https://colab.research.google.com/drive/1bFX0pZczeApeFadrz1AdOb5TDdet2U0Z?usp=sharing</a>!accelerate launch \ --num_cpu_threads_per_process 8 train_network.py \ --network_module=networks.lora \ --pretrained_model_name_or_path="/content/model/nai.ckpt" \ --train_data_dir="/content/drive/MyDrive/sd/dataset/sincos" \ --reg_data_dir="/content/drive/MyDrive/sd/dataset/sincos_reg" \ --output_dir="/content/drive/MyDrive/sd/stable-diffusion-webui/models/lora/" \ --caption_extension=".txt" \ --prior_loss_weight=1 \ --resolution=768 \ --enable_bucket \ --min_bucket_reso=512 \ --max_bucket_reso=1152 \ --train_batch_size=3 \ --learning_rate=1e-4 \ --unet_lr=1.5e-4 \ --text_encoder_lr=1.5e-5 \ # --max_train_steps=8010 \ --mixed_precision="fp16" \ --save_precision="fp16" \ --use_8bit_adam \ --xformers \ --max_train_epochs=8\ --save_every_n_epochs=1\ --save_model_as=safetensors \ --clip_skip=2 \ --seed=23 \ --network_dim=128 \ --network_apha=128 \ --max_token_length=225 \ --cache_latents \ --lr_scheduler="cosine_with_restarts" \ --output_name="minaashido" \ --shuffle_caption \ # --keep_tokens=1 \ # --lr_warmup_steps=$lr_warmup_steps \

Files with SHA256: 343d19f1fd0bd54f371dc3e38fd45f2e1c8c2edab18282176de60b3e744f102f