This is a LoRA I personally use for logo design. After completing a logo design task with it, I decided to share it with everyone.
Actually, it performs quite well in drawing many logos. I trained it with nearly a thousand logo icons, so its generalization ability is better. However, you need to experiment with how to use it. It can draw more than just ducks; in fact, there are no ducks in my training set!
I accidentally discovered that the ducks it draws are very cute, so I decided to share it~
Base model: sd_xl_base_1.0_0.9vae.safetensors
Sampling method: DPM++ 2M Karras
Steps: 22
Resolution: 1024*1024
LoRA strength: 0.85
Trigger words can use "logo" and "brand" to enhance
Description
FAQ
Comments (5)
how many pictures are in your data set and what was your training workflow? I was thinking of training one based on vtuber logos
{
"LoRA_type": "Standard",
"LyCORIS_preset": "full",
"adaptive_noise_scale": 0,
"additional_parameters": "",
"block_alphas": "",
"block_dims": "",
"block_lr_zero_threshold": "",
"bucket_no_upscale": false,
"bucket_reso_steps": 32,
"cache_latents": true,
"cache_latents_to_disk": true,
"caption_dropout_every_n_epochs": 0.0,
"caption_dropout_rate": 0,
"caption_extension": ".txt",
"clip_skip": "1",
"color_aug": false,
"constrain": 0.0,
"conv_alpha": 1,
"conv_block_alphas": "",
"conv_block_dims": "",
"conv_dim": 1,
"debiased_estimation_loss": false,
"decompose_both": false,
"dim_from_weights": false,
"down_lr_weight": "",
"enable_bucket": true,
"epoch": 20,
"factor": -1,
"flip_aug": false,
"fp8_base": false,
"full_bf16": false,
"full_fp16": false,
"gradient_accumulation_steps": 1,
"gradient_checkpointing": true,
"keep_tokens": "0",
"learning_rate": 2.5e-05,
"logging_dir": "",
"lora_network_weights": "",
"lr_scheduler": "cosine",
"lr_scheduler_args": "",
"lr_scheduler_num_cycles": "",
"lr_scheduler_power": "",
"lr_warmup": 0,
"max_bucket_reso": 2048,
"max_data_loader_n_workers": "0",
"max_grad_norm": 1,
"max_resolution": "1024,1024",
"max_timestep": 1000,
"max_token_length": "75",
"max_train_epochs": "",
"max_train_steps": "",
"mem_eff_attn": false,
"mid_lr_weight": "",
"min_bucket_reso": 64,
"min_snr_gamma": 0,
"min_timestep": 0,
"mixed_precision": "fp16",
"model_list": "custom",
"module_dropout": 0,
"multires_noise_discount": 0,
"multires_noise_iterations": 0,
"network_alpha": 16,
"network_dim": 32,
"network_dropout": 0,
"no_token_padding": false,
"noise_offset": 0.1,
"noise_offset_type": "Original",
"num_cpu_threads_per_process": 2,
"optimizer": "Lion",
"optimizer_args": "",
"output_dir": "lora_output",
"output_name": "logo_sdxl",
"persistent_data_loader_workers": false,
"pretrained_model_name_or_path": "sd_xl_base_1.0_0.9vae.safetensors",
"prior_loss_weight": 1.0,
"random_crop": false,
"rank_dropout": 0,
"rank_dropout_scale": false,
"reg_data_dir": "",
"rescaled": false,
"resume": "",
"sample_every_n_epochs": 0,
"sample_every_n_steps": 0,
"sample_prompts": "",
"sample_sampler": "euler_a",
"save_every_n_epochs": 10,
"save_every_n_steps": 0,
"save_last_n_steps": 0,
"save_last_n_steps_state": 1,
"save_model_as": "safetensors",
"save_precision": "fp16",
"save_state": false,
"scale_v_pred_loss_like_noise_pred": false,
"scale_weight_norms": 0,
"sdxl": true,
"sdxl_cache_text_encoder_outputs": false,
"sdxl_no_half_vae": true,
"seed": "12345",
"shuffle_caption": false,
"stop_text_encoder_training_pct": 0,
"text_encoder_lr": 0.0,
"train_batch_size": 5,
"train_data_dir": "logo",
"train_norm": false,
"train_on_input": true,
"training_comment": "",
"unet_lr": 0.0,
"unit": 1,
"up_lr_weight": "",
"use_cp": false,
"use_scalar": false,
"use_tucker": false,
"use_wandb": false,
"v2": false,
"v_parameterization": false,
"v_pred_like_loss": 0,
"vae": "",
"vae_batch_size": 0,
"wandb_api_key": "",
"weighted_captions": false,
"xformers": "xformers"
}
About 800 pictures, the folder 's number is 20, and the other parameters are above.
Ironically people dont spend time reading the description and might have thought this is only about ducks and skipped it.
They're going to miss out on something good.