<h2>Hold up! ⚠</h2><p><strong>This model (v1) is still a work in progress.</strong></p><p>I apologize for the big file, still learning how to train LoRa's efficiently 😉</p><p>Would love some pointers if anyone sees a glaring mistake or error in my first training.</p><h2>The Model</h2><p><strong>Tools:</strong></p><ul><li><p>Custom web scraper for HQ photos off social and Reddit</p></li><li><p><a target="_blank" rel="ugc" href="http://birme.net">birme.net</a> for bulk image cropping</p></li><li><p>Pre-processed w/ BLIP</p></li><li><p>bmaltais'<a target="_blank" rel="ugc" href="http://bmaltais/"> </a><a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">kohya_ss </a>for the actual training</p></li></ul><p></p><p><strong>Parameters:</strong></p><p>121 photos of varying poses and angles, 80% cropped to face, rest to body</p><pre><code>\\ hyoon_v1 parameters

ss_mixed_precision: "bf16",

ss_network_alpha: "128.0",

ss_network_dim: "128",

ss_network_module: "networks.lora",

ss_noise_offset: "None",

ss_num_batches_per_epoch: "6050",

ss_num_epochs: "1",

ss_num_reg_images: "0",

ss_num_train_images: "12100",

ss_optimizer: "torch.optim.adamw.AdamW",

ss_output_name: "hyoon",

ss_prior_loss_weight: "1.0",

ss_resolution: "(672, 672)",

ss_sd_model_name: "runwayml/stable-diffusion-v1-5"</code></pre><p><strong>Recommended Settings:</strong></p><p>I probably messed up somewhere, because the best results I have gotten so far are from the following triggers:</p><p><code><lora:hyoon_v1:0.8></code> <code>(hyoon:1.2)</code></p><img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/9574e3b8-2261-400b-1200-13da15538e00/width=525/9574e3b8-2261-400b-1200-13da15538e00" /><p>Prompt used:</p><p><code><lora:hyoon_v1:0.8>,(8k, RAW portrait,photorealistic,masterpiece:1.2), (full body) portrait of (hyoon:1.2) posting on a grassy hill, detailed cute face, medium breasts, low-cut shirt, cleavage, slender body, looking at the viewer, smiling, <lora:fashionGirl_v40:0.1></code></p><p><code><br />Negative prompt: easynegative:0.4, illustration, painting, drawing, art, sketch, deformed, ugly, mutilated</code></p><p>Model used: <strong>doctorDiffusionV08</strong></p><p><br /><strong>WARNING: </strong>this model seems to be on the naughtier side, so please respect the person and use negative prompts - <code>nude, naked, nipples, *****</code></p><p>Very oddly, <strong>Euler a </strong>seems to be the only 'particularly horny' sampler. No idea why or how that happens, but please feel free to enlighten me.</p><p><strong>DISCLAIMER: </strong>Please respect the person of the model, use common sense, and don't get in trouble. What you do during fap time is your private business, but don't ruin it for everyone else by being reckless and damaging the reputation of the person or the website and the great services they are providing.</p>