HiDream-I1 FP8

UPDATE 4/11/2025

Lots of changes here. First off, I've cleaned up the code and added caching of the quantized transformer so that it's only really slow on the first run in a given mode (full, dev, or fast). On subsequent runs, the quantized transformer will be loaded from disk. Also, I've changed the llama model location so that it links to an ablated, pre-quantized version, which loads significantly faster.

I've added a gradio interface. After installing with:

pip install -r requirements.txt

run with:

python gradio_torchao.py

Like the command line version, the gradio interface caches the transformer and will run faster on subsequent prompts, particularly if you keep using the same mode rather than switching. It also comes with a "negative prompt scale" parameter that allows you to adjust the strength of the negative prompt. Negative prompts even work on the CFG-free dev and fast models because I've figured out a trick to combine the positive and negative embeddings with subtraction rather than concatenation, although the negative prompt acts a bit differently for dev and fast, so you'll want to do some experimentation.

Original HiDream-I1 FP8 readme

This works on my 4090, using around 18 gigs of vram. It's slow as hell because it's quantizing the models on load, but it's a proof of concept and it works.

This code is garbage. Google AI Studio and ChatGPT o3-mini-high weren't really up to the task, which I ultimately figured out myself, but because of that the code probably has way more cruft than it should.

Install the requirements with:

pip install -r requirements.txt

Run with:

python inference_torchao.py --prompt "An avocado in the shape of a chair, or something."

or

accelerate launch inference_torchao.py --prompt "A woman lying in the grass."

Things I should do later but probably won't because ComfyUI will support it natively by the time I get around to it:

• Make it load models that are already quantized so it's not so excruciatingly slow to run
• Figure out some way to swap llama to the CPU rather than unloading it, so it'll be viable to keep everything loaded for multiple inferences
• Add a --seed command line parameter (this should be easy)
• Experiment with FP6 and see if I can make it fit under 16 gigs.
• Fix the stupid --resolution command line parameter.

HiDream-I1

HiDream-I1 is a new open-source image generative foundation model with 17B parameters that achieves state-of-the-art image generation quality within seconds.

For more features and to experience the full capabilities of our product, please visit https://vivago.ai/.

Project Updates

• 🤗 April 8, 2025: We've launched a Hugging Face Space for HiDream-I1-Dev. Experience our model firsthand at https://huggingface.co/spaces/HiDream-ai/HiDream-I1-Dev!
• 🚀 April 7, 2025: We've open-sourced the text-to-image model HiDream-I1.

Models

We offer both the full version and distilled models. For more information about the models, please refer to the link under Usage.

Name	Script	Inference Steps	HuggingFace repo
HiDream-I1-Full	inference.py	50	🤗 HiDream-I1-Full
HiDream-I1-Dev	inference.py	28	🤗 HiDream-I1-Dev
HiDream-I1-Fast	inference.py	16	🤗 HiDream-I1-Fast

Quick Start

Please make sure you have installed Flash Attention. We recommend CUDA versions 12.4 for the manual installation.

pip install -r requirements.txt
pip install -U flash-attn --no-build-isolation

Then you can run the inference scripts to generate images:

# For full model inference
python ./inference.py --model_type full

# For distilled dev model inference
python ./inference.py --model_type dev

# For distilled fast model inference
python ./inference.py --model_type fast

Note

The inference script will try to automatically download meta-llama/Llama-3.1-8B-Instruct model files. You need to agree to the license of the Llama model on your HuggingFace account and login using huggingface-cli login in order to use the automatic downloader.

Gradio Demo

We also provide a Gradio demo for interactive image generation. You can run the demo with:

python gradio_demo.py 

Evaluation Metrics

DPG-Bench

Model	Overall	Global	Entity	Attribute	Relation	Other
PixArt-alpha	71.11	74.97	79.32	78.60	82.57	76.96
SDXL	74.65	83.27	82.43	80.91	86.76	80.41
DALL-E 3	83.50	90.97	89.61	88.39	90.58	89.83
Flux.1-dev	83.79	85.80	86.79	89.98	90.04	89.90
SD3-Medium	84.08	87.90	91.01	88.83	80.70	88.68
Janus-Pro-7B	84.19	86.90	88.90	89.40	89.32	89.48
CogView4-6B	85.13	83.85	90.35	91.17	91.14	87.29
HiDream-I1	85.89	76.44	90.22	89.48	93.74	91.83

GenEval

Model	Overall	Single Obj.	Two Obj.	Counting	Colors	Position	Color attribution
SDXL	0.55	0.98	0.74	0.39	0.85	0.15	0.23
PixArt-alpha	0.48	0.98	0.50	0.44	0.80	0.08	0.07
Flux.1-dev	0.66	0.98	0.79	0.73	0.77	0.22	0.45
DALL-E 3	0.67	0.96	0.87	0.47	0.83	0.43	0.45
CogView4-6B	0.73	0.99	0.86	0.66	0.79	0.48	0.58
SD3-Medium	0.74	0.99	0.94	0.72	0.89	0.33	0.60
Janus-Pro-7B	0.80	0.99	0.89	0.59	0.90	0.79	0.66
HiDream-I1	0.83	1.00	0.98	0.79	0.91	0.60	0.72

HPSv2.1 benchmark

Model	Averaged	Animation	Concept-art	Painting	Photo
Stable Diffusion v2.0	26.38	27.09	26.02	25.68	26.73
Midjourney V6	30.29	32.02	30.29	29.74	29.10
SDXL	30.64	32.84	31.36	30.86	27.48
Dall-E3	31.44	32.39	31.09	31.18	31.09
SD3	31.53	32.60	31.82	32.06	29.62
Midjourney V5	32.33	34.05	32.47	32.24	30.56
CogView4-6B	32.31	33.23	32.60	32.89	30.52
Flux.1-dev	32.47	33.87	32.27	32.62	31.11
stable cascade	32.95	34.58	33.13	33.29	30.78
HiDream-I1	33.82	35.05	33.74	33.88	32.61

License

The code in this repository and the HiDream-I1 models are licensed under MIT License.