File size: 588 Bytes
c5d43bc |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 |
#!/usr/bin/env python3
import torch
from diffusers import ConsistencyModelPipeline, UNet2DModel
device = "cpu"
# Load the cd_bedroom256_lpips checkpoint.
model_id_or_path = "openai/diffusers-cd_bedroom256_lpips"
pipe = ConsistencyModelPipeline.from_pretrained(model_id_or_path)
pipe.to(device)
# Multistep sampling
# Timesteps can be explicitly specified; the particular timesteps below are from the original Github repo:
# https://github.com/openai/consistency_models/blob/main/scripts/launch.sh#L83
for _ in range(10):
image = pipe(timesteps=[17, 0]).images[0]
image.show()
|