Total runs:
77.1K
Run Growth:
-24.6K
Growth Rate:
-31.32%
Sponsored by
APIMart
HiDream-ai / Hi-MAR
Model's Last Updated: May 24 2025
Introduction of Hi-MAR
Model Details of Hi-MAR
Hi-MAR
🖥️
GitHub
｜ 🌐
Project Page
| 🤗
Hugging Face
| 📑
Paper
| 📖
PDF
Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots (ICML 2025)
This is the official repository for the Paper "Hierarchical Masked Autoregressive Models with Low-Resolution Token Pivots"
Overview
We present a Hierarchical Masked Autoregressive models (Hi-MAR) that pivot on low-resolution image tokens to trigger hierarchical autoregressive modeling in a multi-phase manner.
🔍 What We're Working to Solve?
- Incapable of utilizing global context in early-stage predictions of the next-token paradigm
- Training-inference discrepancy across multi-scale predictions
- Suboptimal multi-scale probability distribution modeling
- Lack of global information in the denoising process of the MLP-based Diffusion head
🔥 Updates
- [2025.05.22] Upload inference code and pretrained class-conditional Hi-MAR models trained on ImageNet 256x256.
🏃🏼 Inference
Environment Requirement
Clone the repo:
git clone https://github.com/HiDream-ai/himar.git
cd himar
Install dependencies:
conda env create -f environment.yaml
conda activate himar
Model Download
Download VAE from the link in the MAR Github .
You can download our pre-trained Hi-MAR models directly from the links provided here.
Evaluation
Evaluate Hi-MAR-B on ImageNet256x256:
torchrun --nproc_per_node=8 --nnodes=1 main_himar.py --img_size 256 --vae_path /path/to/vae --vae_embed_dim 16 --vae_stride 16 --patch_size 1 --model himar_base --diffloss_d 6 --diffloss_w 1024 --output_dir ./himar_base_test --resume /path/to/Hi-MAR-B --num_images 50000 --num_iter 4 --cfg 2.5 --re_cfg 2.7 --cfg_schedule linear --cond_scale 8 --cond_dim 16 --two_diffloss --global_dm --gdm_d 6 --gdm_w 512 --eval_bsz 256 --load_epoch -1 --head 8 --ratio 4 --cos --evaluate
Evaluate Hi-MAR-L on ImageNet256x256:
torchrun --nproc_per_node=8 --nnodes=1 main_himar.py --img_size 256 --vae_path /path/to/vae --vae_embed_dim 16 --vae_stride 16 --patch_size 1 --model himar_large --diffloss_d 8 --diffloss_w 1280 --output_dir ./himar_large_test --resume /path/to/Hi-MAR-L --num_images 50000 --num_iter 4 --cfg 3.5 --re_cfg 3.5 --cfg_schedule linear --cond_scale 8 --cond_dim 16 --two_diffloss --global_dm --gdm_d 8 --gdm_w 512 --eval_bsz 256 --load_epoch -1 --head 8 --ratio 4 --cos --evaluate
Evaluate Hi-MAR-H on ImageNet256x256:
torchrun --nproc_per_node=8 --nnodes=1 main_himar.py --img_size 256 --vae_path /path/to/vae --vae_embed_dim 16 --vae_stride 16 --patch_size 1 --model himar_huge --diffloss_d 12 --diffloss_w 1536 --output_dir ./himar_huge_test --resume /path/to/Hi-MAR-H --num_images 50000 --num_iter 12 --cfg 3.2 --re_cfg 5.5 --cfg_schedule linear --cond_scale 8 --cond_dim 16 --two_diffloss --global_dm --gdm_d 12 --gdm_w 768 --eval_bsz 256 --load_epoch -1 --head 12 --ratio 4 --cos --evaluate
🌟 Star and Citation
If you find our work helpful for your research, please consider giving a star⭐ on this repository and citing our work.
💖 Acknowledgement
Thanks to the contribution of MAR
Runs of HiDream-ai Hi-MAR on huggingface.co
0
Total runs
0
24-hour runs
0
3-day runs
0
7-day runs
0
30-day runs
More Information About Hi-MAR huggingface.co Model
Url of Hi-MAR
Hi-MAR huggingface.co Url
Provider of Hi-MAR huggingface.co
Other API from HiDream-ai
Updated:June 17 2025
Total runs:
9.6K
Run Growth:
2.8K
Growth Rate:
28.46%
Total runs:
2.7K
Run Growth:
947
Growth Rate:
35.01%
Total runs:
788
Run Growth:
-6.4K
Growth Rate:
-818.33%
Total runs:
523
Run Growth:
-560
Growth Rate:
-104.87%
Total runs:
319
Run Growth:
-235
Growth Rate:
-82.46%