noelle-krea2-lora

Krea 2 character LoRA β€” trigger no3l woman. Trained on Krea 2 RAW, intended for inference on Krea 2 Turbo.

Status: βœ… training complete

Training configuration

Field Value
Base model (train) krea/Krea-2-Raw (raw.safetensors, 12B DiT, undistilled)
Inference target krea/Krea-2-Turbo (8-step distilled) β€” RAW-train / Turbo-infer
VAE Qwen-Image VAE (qwen_image_vae.safetensors)
Text encoder Qwen3-VL-4B (qwen3vl_4b_bf16.safetensors)
Trainer Musubi Tuner (kohya-ss), networks.lora_krea2
LoRA rank/dim 24, alpha 24, all Linear layers of the DiT
Optimizer adamw8bit, lr 1e-4
Precision bf16, gradient checkpointing, SDPA
Timestep sampling krea2_shift (resolution-aware), weighting_scheme none
Trigger no3l woman (golden tan + blonde hair (high bun) + blue eyes + face baked into the trigger; white bikini + gold jewelry described)
Dataset 64 images + paired NL captions (seductive beach/bikini set, Seedream-generated, Dad Bod character #73); white bandeau bikini + gold jewelry + pose/expression/framing/background described, identity in trigger. batch 1, 1024 multi-res buckets, num_repeats 2
Schedule 12 epochs = 1536 steps, save every 2 epochs, seed 42
Hardware 1x NVIDIA H100 80GB (RunPod, EU-NL-1, Musubi volume)

Loss per checkpoint

avr_loss is a rolling average β€” flow-matching loss is near-flat, so use it as a sanity check and pick the best checkpoint visually.

checkpoint epoch step avr_loss
-000002 2 256 0.0460
-000004 4 512 0.0433
-000006 6 768 0.0396
-000008 8 1024 0.0435
-000010 10 1280 0.0407
final 12 1536 0.0397

min avr_loss 0.0368 Β· final 0.0397

Inference (Krea 2 Turbo, Musubi Tuner)

Prepend the trigger no3l woman to the prompt. Recommended: stack over the house style LoRA.

python src/musubi_tuner/krea2_generate_image.py \
    "no3l woman, <scene>" \
    --dit turbo.safetensors --vae qwen_image_vae.safetensors \
    --text_encoder qwen3vl_4b_bf16.safetensors \
    --steps 8 --guidance_scale 1 --mu 1.15 --width 1024 --height 1280 \
    --attn_mode torch --lora_weight <checkpoint>.safetensors --lora_multiplier 1.0

Dataset samples

Representative image+caption pairs from the training set:

Noelle_DB88-portrait-001-seedream.jpg

Noelle_DB88-portrait-001-seedream

no3l woman, close-up shot framing her face and upper chest, wearing a white bandeau bikini top with a gold ring centerpiece, gold hoop earrings visible, direct gaze with lips slightly parted, plain white background.

Noelle_DB88-medium-002-seedream.jpg

Noelle_DB88-medium-002-seedream

no3l woman, medium shot, body turned three-quarter with one hand resting on her hip, wearing a white bandeau bikini top with a gold ring centerpiece and a white tie-side bikini bottom with gold beaded charms, gold hoop earrings, a confident smirk with a direct gaze, plain dusty rose background.


Auto-generated by training watchdog.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for Zaytron40k/noelle-krea2-lora

Base model

krea/Krea-2-Raw
Adapter
(1,300)
this model