Skip to content

hidream_o1_transformer.md

Latest commit

 

History

History
34 lines (23 loc) · 1.2 KB

hidream_o1_transformer.md

File metadata and controls

34 lines (23 loc) · 1.2 KB

HiDreamO1Transformer2DModel

A Qwen3-VL based raw pixel patch transformer for HiDream-O1-Image.

HiDream-O1 does not use a VAE. The transformer predicts raw RGB pixel patches through the O1 denoising path added on top of Qwen3-VL.

The model can be loaded with the following code snippet.

import torch
from diffusers import HiDreamO1Transformer2DModel

transformer = HiDreamO1Transformer2DModel.from_pretrained(
    "HiDream-ai/HiDream-O1-Image",
    torch_dtype=torch.bfloat16,
)

HiDreamO1Transformer2DModel

[[autodoc]] HiDreamO1Transformer2DModel