import torch
from diffusers import DiffusionPipeline
# switch to "mps" for apple devices
pipe = DiffusionPipeline.from_pretrained("MickJ/Z-Image-Turbo-fp8", dtype=torch.bfloat16, device_map="cuda")
prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k"
image = pipe(prompt).images[0]an fp8 version of Tongyi-MAI/Z-Image-Turbo, converted with the checkpoint of the original transformer component with:
python -m sglang.multimodal_gen.tools.convert_hf_to_fp8 \
--model-dir /root/.cache/huggingface/hub/models--Tongyi-MAI--Z-Image-Turbo/snapshots/f332072aa78be7aecdf3ee76d5c247082da564a6/transformer/
--save-dir /root/.cache/huggingface/hub/models--Tongyi-MAI--Z-Image-Turbo-fp8/snapshots/f332072aa78be7aecdf3ee76d5c247082da564a6/transformer/
For SGLang-Diffusion CI usage, guard the --transformer-path server arg feature
- Downloads last month
- 3,089
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support