import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("PonyMeng/TPSO", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]TPSO Unconditional Contexts
Precomputed unconditional contexts for TPSO v0.1.0. This repository does not contain diffusion-model weights.
| Model | File |
|---|---|
| Stable Diffusion 1.5 | sd15_kappa0.8_lambda1.pt |
| Stable Diffusion 2.1 | sd21_kappa0.8_lambda1.pt |
| Stable Diffusion 3.5 Medium | sd35_kappa0.8_lambda0.pt |
TPSO downloads and verifies the matching file automatically:
tpso-generate \
--model sd15 \
--prompt "A photograph of a red panda in a bamboo forest" \
--num-images 4
Each file contains optimized CLIP representations and metadata. The TPSO
loader uses torch.load(..., weights_only=True) and verifies its SHA-256 digest.
TPSO code is Apache-2.0. Upstream model licenses and access terms still apply.
- Downloads last month
- -