krea2-style-v1

Krea 2 style LoRA (digital art). Trained on Krea 2 RAW, intended for inference on Krea 2 Turbo.

Status: ✅ 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) — RAW-train / Turbo-infer
VAE	Qwen-Image VAE (qwen_image_vae.safetensors)
Text encoder	Qwen3-VL-4B-Instruct (qwen3vl_4b_bf16.safetensors)
Trainer	Musubi Tuner (kohya-ss), networks.lora_krea2
LoRA	rank/dim 32, alpha 32, all 264 Linear layers of the DiT
Optimizer	adamw8bit, lr 1e-4
Precision	bf16, gradient checkpointing, SDPA
Timestep sampling	krea2_shift (resolution-aware), weighting_scheme none
Dataset	337 images + paired captions, batch 4, 1024 multi-res buckets, num_repeats 1
Schedule	40 epochs = 3960 steps, save every 5 epochs, seed 42
Hardware	1x NVIDIA H100 80GB (RunPod)

Loss per checkpoint

avr_loss is a rolling average (flow-matching loss is near-flat; use it as a sanity signal, pick the best checkpoint visually).

checkpoint	epoch	step	avr_loss
-000005	5	495	0.0774
-000010	10	990	0.0782
-000015	15	1485	0.0744
-000020	20	1980	0.0762
-000025	25	2475	0.0750
-000030	30	2970	0.0750
-000035	35	3465	0.0747
-000040 (final)	40	3960	0.0735

Inference (Krea 2 Turbo, Musubi Tuner)

python src/musubi_tuner/krea2_generate_image.py \
    "your prompt" \
    --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 1024 \
    --attn_mode torch --lora_weight <checkpoint>.safetensors --lora_multiplier 1.0

Dataset samples

Two representative image+caption pairs from the 337-image training set:

000233.jpeg

A high-angle view of an empty office workspace. A black rolling chair faces a light wooden desk with two monitors. One screen displays various charts, while the other shows text. A black keyboard, mouse, closed notebook, and a small potted plant rest on the brightly lit surface.

0004.jpeg

A medium shot shows a man and woman facing each other in profile at a cafe table by a window. The brown-haired man in a dark sweater drinks from a white mug. The black-haired woman in a red top rests her chin on clasped hands. Daylight illuminates the wooden table.

---

Auto-generated by training watchdog.