Views
No views yet
cfg (beta = 1/cfg).alpha.time_embedding.cond_proj
(time_cond_proj_dim = 2 x 512 = 1024):
timestep_cond = concat([gse(1/beta - 1), gse(1/alpha - 1)]).ema_decay 0.9999) from checkpoint-75000.| item | value |
|---|---|
| init | sd-legacy/stable-diffusion-v1-5 UNet (cond_proj zero-init) |
| data | LAION Aesthetics 5+ subset, 19.1M images, WebDataset |
| steps | 75,000 (fp16, global batch 256, lr 1e-5 constant + 1k warmup) |
| objective | dual-guidance distillation (--rm_guidance --sg_guidance), min_alpha 0.25, mixup_ratio 0.2 |
| alpha range | trained on alpha ∈ [0.25, 1.0] — evaluating below 0.25 is extrapolation |
StableDiffusionPipeline cannot express the dual (rm, sg) conditioning —
use the custom denoise loop (sample_sg_fullft.py::sg_generate in the training repo),
which does a single UNet forward per step (no CFG doubling):1from diffusers import UNet2DConditionModel
2unet = UNet2DConditionModel.from_pretrained("jhtwosuin/sd15-sg-fullft-75k")
3# vae / text_encoder / tokenizer / scheduler come frozen from sd-legacy/stable-diffusion-v1-5
4# timestep_cond = cat([gse(cfg-1), gse(1/alpha-1)]) ; recommended cfg 7.5, alpha ∈ {0.8, 0.5, 0.25}