gatilin/hidream_o1_vision_pixel_io

HiDream-O1 visual branch and native pixel input/output interfaces.

Preserves Qwen3-VL visual/deepstack modules, patch-32 bottleneck pixel input, timestep embedder and pixel output projection. The shared 8B-class language/flow backbone is not included. This checkpoint retains a visual encoder despite VAE-free pixel generation. Full visual question-answering quality was not verified and is not claimed. Prompt Agent weights are not included.

Source and conversion

Model/weights: HiDream-ai/HiDream-O1-Image, fixed revision 0b0901d99f200389e138c61946af1185f5f49a13. Native code: HiDream-ai/HiDream-O1-Image, fixed revision 2c2d29ff729e48f33e41f49edfdbd81d5ac103b4.

All selected tensor names, shapes, dtypes and payload bytes are preserved, except the explicit IBQ EMA shadow-name mapping. See extraction.json, provenance.json and tensor/conversion mappings. This repository contains component weights and a local native loader, not an AutoModel.from_pretrained checkpoint.

Load

Install requirements.txt in a compatible environment, then:

from huggingface_hub import snapshot_download
import sys
import torch
root = snapshot_download("gatilin/hidream_o1_vision_pixel_io")
sys.path.insert(0, root)
from component import load_model
model = load_model(root, dtype=torch.float32, device="cpu")

The model is returned in evaluation mode. Run inference under torch.inference_mode(). Inputs are native tensors/patches, not arbitrary PIL images; inspect component.py and the included validation script for the exact layout. CPU float32 was tested; CUDA/MPS and full end-to-end benchmarks were not tested.

Validation

See validation.json for the executed component tests, shapes and limitations. Source payload equality and native strict loading are required before release. Component tests do not establish full-model generation or understanding quality. Original stochastic decoding settings and normalization conventions must be retained.

License and attribution

Original upstream notices and license files are retained. Reused third-party components keep their original terms. This extraction does not claim authorship of the underlying trained model.