Heouzen/LoKR_mc_woman_FLUX1D

LoKRmcwoman_FLUX1D

Model trained with AI Toolkit by Ostris <Gallery />

LoKR mc_woman Training Logs

Berikut adalah rincian progres training berdasarkan strategi dataset yang diterapkan:

Strategy Breakdown:

• —Phase 1 (0-3000): Menggunakan seluruh dataset untuk membangun fondasi anatomi yang kuat.
• —Phase 2 (3000-3500): Penyaringan dataset (remove setengah) dilakukan untuk menghilangkan "noise" agar model tidak bingung dengan wajah yang kurang mirip.
• —Phase 3 (3500-4000): Fokus pada close-up untuk memaksa model mempelajari tekstur mikro kulit (pori-pori dan kerutan halus).

Rekomendasi Penggunaan:

Untuk hasil paling natural, disarankan menggunakan Checkpoint 2500 atau 3600 dengan strength LoRA di kisaran 0.8 - 0.9 dan guidance scale 3.5.

training_summary:

• —step: 2000 loss: 0.647265 status: "🟢 BEST - Phase 1: Default Datasets"
• —step: 2100 loss: 0.421459 status: "🔵 STABLE"
• —step: 2200 loss: 0.385189 status: "🔵 STABLE"
• —step: 2300 loss: 0.300921 status: "🔵 STABLE"
• —step: 2400 loss: 0.594884 status: "🔵 STABLE"
• —step: 2500 loss: 0.328804 status: "🟢 BEST - Phase 1: Locked-in Texture"
• —step: 2600 loss: 0.561934 status: "🔵 STABLE"
• —step: 2700 loss: 0.639280 status: "🔵 STABLE"
• —step: 2800 loss: 0.519834 status: "🔵 STABLE"
• —step: 2900 loss: 0.599700 status: "🔵 STABLE"
• —step: 3100 loss: 0.476988 status: "🟡 REFINED - Phase 2: Half Dataset Removed"
• —step: 3200 loss: 0.525119 status: "🟡 REFINED"
• —step: 3300 loss: 0.635466 status: "🟡 REFINED"
• —step: 3400 loss: 0.475486 status: "🟡 REFINED"
• —step: 3600 loss: 0.307630 status: "🟢 BEST - Phase 3: Close-up Face Focus"
• —step: 3700 loss: 0.566942 status: "🟠 FOCUSED"
• —step: 3800 loss: 0.427737 status: "🟠 FOCUSED"
• —step: 3900 loss: 0.571653 status: "🟠 FOCUSED"

PENING PALAK

Trigger words

You should use mc_woman to trigger the image generation.

Download model and use it with ComfyUI, AUTOMATIC1111, SD.Next, Invoke AI, etc.

Weights for this model are available in Safetensors format.

Download them in the Files & versions tab.

Use it with the 🧨 diffusers library

from diffusers import AutoPipelineForText2Image
import torch

pipeline = AutoPipelineForText2Image.from_pretrained('black-forest-labs/FLUX.1-dev', torch_dtype=torch.bfloat16).to('cuda')
pipeline.load_lora_weights('Heouzen/LoKR_mc_woman_FLUX1D', weight_name='LoKR_mc_woman_FLUX1D.safetensors')
image = pipeline('mc_woman,  a young woman with long,  wavy dark brown hair,  is seated on a brown sofa. She has a warm expression,  looking slightly to her left with a gentle smile. She is dressed in a white short-sleeved,  form-fitting crop top that reveals her midriff,  paired with a light blue denim mini skirt. A black watch with a round face is on her left wrist,  and her left hand rests casually on her right thigh. Behind her is a wall with textured glass block panels and a round mirror.').images[0]
image.save("my_image.png")

For more details, including weighting, merging and fusing LoRAs, check the documentation on loading LoRAs in diffusers