Downloads · 30 days
0
sd15-dpo-uk-real-b2000
fassabilf/sd15-dpo-uk-real-b2000
sd15-dpo-uk-real-b2000 is a machine learning model from fassabilf. Use it for the machine learning task on the model card, and read the license before you ship it in a product. It is set up for diffusers. The card lists the license as creativeml-openrail-m.
Run diagnostik, bukan arm tabel utama. Setelan paper Diffusion-DPO (lr 1e-8 x --scalelr = 3,2e-7 efektif, warmup 25 % = 236 step, 15 epoch / 945 step, --betadpo 2000), dilatih atas pasangan preferensi dpo/ di dataset…
Access
Public
Updated Sep 23, 2026
Repo size
24.1 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors24.1 GB · 100%
At a glance
- Library
- diffusers
- License
- creativeml-openrail-m
- Access
- Public
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026
- SHA
- d2d15888
Model card
From the Hugging Face model README
SD 1.5 Diffusion-DPO foto asli UK — run pembanding beta=2000, setelan lr paper
Run diagnostik, bukan arm tabel utama. Setelan paper Diffusion-DPO (lr 1e-8 x --scale_lr
= 3,2e-7 efektif, warmup 25 % = 236 step, 15 epoch / 945 step, --beta_dpo 2000), dilatih atas
pasangan preferensi dpo/ di dataset fassabilf/ift-train-uk-real (2.000 train / 600 val,
20 okupasi UK), base stable-diffusion-v1-5/stable-diffusion-v1-5, seed 0, 63 step/epoch.
Pembandingnya: fassabilf/sd15-dpo-uk-real (beta=5000, lr 1e-5 setelan proyek).
unet_ep{3..15} = checkpoint-{189..945}/unet; ep3–14 fp16 hasil compact, ep15 fp32 utuh.
logs/ berisi tensorboard run ini. Analisis dan pembacaannya ada di repo kode
(analysis/dpo_uk_b2000/COMPARE.md): loss DPO turun monoton 0,693 -> 0,615 sementara MAE_real
test ep15 memburuk +0,0195 (CI [+0,0033, +0,0386]) vs base SD 1.5 UK.
Diunggah belakangan (2026-09-23) semata sebagai cadangan sebelum checkpoint dihapus dari pod.
- Library
- diffusers
- License
- creativeml-openrail-m
- Created
- Sep 23, 2026
- Updated
- Sep 23, 2026