<h3 id="heading-2">Updated</h3>Used PonyXL to train this one. Updated dataset to achieve better performance. It enhanced this concept on PonyXL and related models. See the examples first.I should say the Pony model fixed many problems I met. I do advice people to give it a try.<h3 id="heading-634">Introduction</h3>An experimental LyCORIS to generate futanari filling condom. Use the key words to control the size of penis and condom.Still a lot of room for improvement.<h3 id="heading-230">Pros</h3><ul><li>Can be combined with multiple Loras</li><li>Little impact on the art style</li></ul><h3 id="heading-97">Cons</h3><ul><li>Using 3D materials for training, the art style may become strange under high weights</li><li>Twisted penises and multiple penises</li><li>Bad glan shape</li><li>Lots of trials needed</li></ul><h3 id="heading-635">How to use it</h3>I recommand using it in weight 1-1.2.<h3 id="heading-636">Training details</h3>Traingset is of about 107 images. Half of them are chosen as regularization images. Total steps are about 7600.resolution=768*768batch_size=2epoch=10network_dim=32network_alpha=16conv_dim=8conv_alpha=4clip_skip=2Using Prodigy as optimizer, setting as the author recommend.

Condom filling (No Update)

Used PonyXL to train this one. Updated dataset to achieve better performance. It enhanced this concept on PonyXL and related models. See the examples first.I should say the Pony model fixed many problems I met. I do advice people to give it a try.Parameters:{
 "LoRA_type": "LyCORIS/LoCon",
 "LyCORIS_preset": "full",
 "adaptive_noise_scale": 0,
 "additional_parameters": "",
 "block_alphas": "",
 "block_dims": "",
 "block_lr_zero_threshold": "",
 "bucket_no_upscale": true,
 "bucket_reso_steps": 128,
 "bypass_mode": false,
 "cache_latents": false,
 "cache_latents_to_disk": false,
 "caption_dropout_every_n_epochs": 0.0,
 "caption_dropout_rate": 0,
 "caption_extension": ".txt",
 "clip_skip": 2,
 "color_aug": false,
 "constrain": 0.0,
 "conv_alpha": 4,
 "conv_block_alphas": "",
 "conv_block_dims": "",
 "conv_dim": 8,
 "dataset_config": "",
 "debiased_estimation_loss": false,
 "decompose_both": false,
 "dim_from_weights": false,
 "dora_wd": false,
 "down_lr_weight": "",
 "enable_bucket": true,
 "epoch": 10,
 "factor": -1,
 "flip_aug": false,
 "fp8_base": false,
 "full_bf16": false,
 "full_fp16": false,
 "gpu_ids": "",
 "gradient_accumulation_steps": "1",
 "gradient_checkpointing": true,
 "keep_tokens": 1,
 "learning_rate": 1.0,
 "log_tracker_config": "",
 "log_tracker_name": "",
 "logging_dir": "/root/autodl-tmp/output",
 "lora_network_weights": "",
 "lr_scheduler": "cosine",
 "lr_scheduler_args": "",
 "lr_scheduler_num_cycles": "",
 "lr_scheduler_power": "",
 "lr_warmup": 0,
 "max_bucket_reso": 4096,
 "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": 1024,
 "min_snr_gamma": 0,
 "min_timestep": 0,
 "mixed_precision": "bf16",
 "model_list": "custom",
 "module_dropout": 0,
 "multi_gpu": false,
 "multires_noise_discount": 0.1,
 "multires_noise_iterations": 6,
 "network_alpha": 8,
 "network_dim": 16,
 "network_dropout": 0.1,
 "noise_offset": 0,
 "noise_offset_type": "Multires",
 "num_cpu_threads_per_process": 2,
 "num_machines": 1,
 "num_processes": 1,
 "optimizer": "Prodigy",
 "optimizer_args": "weight_decay=0.01 d_coef=1 use_bias_correction=True safeguard_warmup=False betas=0.9,0.99 decouple=True ",
 "output_dir": "/root/autodl-tmp/output",
 "output_name": "condom_filling-dropout",
 "persistent_data_loader_workers": false,
 "pretrained_model_name_or_path": "/root/kohya_ss/models/ponyDiffusionV6XL.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": 1,
 "sample_every_n_steps": 0,
 "sample_prompts": "score_9, score_8_up, score_7_up, anime BREAK, hyper_coneja, 1girl, solo, long hair, breasts, open mouth, black hair, thighhighs, white background, navel, nipples, purple eyes, nude, penis, dark skin, spread legs, armpits, huge breasts, arms up, white thighhighs, dark-skinned female, uncensored, erection, squatting, testicles, arms behind head, condom, veins, futanari, veiny penis, large penis, used condom, huge penis, condom on penis, large testicles, horse penis --n source_pony, source_furry, source_cartoon, greyscale, from below --w 1024 --h 1024\n ",
 "sample_sampler": "euler_a",
 "save_every_n_epochs": 1,
 "save_every_n_steps": 0,
 "save_last_n_steps": 0,
 "save_last_n_steps_state": 0,
 "save_model_as": "safetensors",
 "save_precision": "bf16",
 "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": "31337",
 "shuffle_caption": true,
 "stop_text_encoder_training_pct": 0,
 "text_encoder_lr": 1.0,
 "train_batch_size": 5,
 "train_data_dir": "/root/autodl-tmp/condom_filling",
 "train_norm": false,
 "train_on_input": true,
 "training_comment": "",
 "unet_lr": 1.0,
 "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": "/root/kohya_ss/models/sdxl_vae_fp16fix.safetensors",
 "vae_batch_size": 0,
 "wandb_api_key": "",
 "wandb_run_name": "",
 "weighted_captions": false,
 "xformers": "xformers"
}<pre><code></code></pre>

condom_filling-dropout.safetensors

PonyXL

v1.0

From Original Post:UpdatedUsed PonyXL to train this one. Updated dataset to achieve better performance. It enhanced this concept on PonyXL and related models. See the examples first.I should say the Pony model fixed many problems I met. I do advice people to give it a try.IntroductionAn experimental LyCORIS to generate futanari filling condom. Use the key words to control the size of penis and condom.Still a lot of room for improvement.Pros<ul><li>Can be combined with multiple Loras</li><li>Little impact on the art style</li></ul>Cons<ul><li>Using 3D materials for training, the art style may become strange under high weights</li><li>Twisted penises and multiple penises</li><li>Bad glan shape</li><li>Lots of trials needed</li></ul>

Condom filling

Used PonyXL to train this one. Updated dataset to achieve better performance. It enhanced this concept on PonyXL and related models. See the examples first.

I should say the Pony model fixed many problems I met. I do advice people to give it a try.



Files with SHA256: fad43778cd983d24f523be803c65bf6bf4aedd9f1be1cd2428e4e8fbe59cf9e0