Hunyuan-DiT : A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding
This repo contains PyTorch model definitions, pre-trained weights and inference/sampling code for our paper exploring Hunyuan-DiT. You can find more visualizations on our project page.
DialogGen: Multi-modal Interactive Dialogue System for Multi-turn Text-to-Image Generation
- Jun 19, 2024: 🎉 ControlNet is released, supporting canny, pose and depth control. See training/inference codes for details.
- Jun 13, 2024: ⚡ HYDiT-v1.1 version is released, which mitigates the issue of image oversaturation and alleviates the watermark issue. Please check HunyuanDiT-v1.1 and Distillation-v1.1 for more details.
- Jun 13, 2024: 🚚 The training code is released, offering full-parameter training and LoRA training.
- Jun 06, 2024: 🎉 Hunyuan-DiT is now available in ComfyUI. Please check ComfyUI for more details.
- Jun 06, 2024: 🚀 We introduce Distillation version for Hunyuan-DiT acceleration, which achieves 50% acceleration on NVIDIA GPUs. Please check Distillation for more details.
- Jun 05, 2024: 🤗 Hunyuan-DiT is now available in 🤗 Diffusers! Please check the example below.
- Jun 04, 2024: 🌐 Support Tencent Cloud links to download the pretrained models! Please check the links below.
- May 22, 2024: 🚀 We introduce TensorRT version for Hunyuan-DiT acceleration, which achieves 47% acceleration on NVIDIA GPUs. Please check TensorRT-libs for instructions.
- May 22, 2024: 💬 We support demo running multi-turn text2image generation now. Please check the script below.
Welcome to our web-based Tencent Hunyuan Bot, where you can explore our innovative products! Just input the suggested prompts below or any other imaginative prompts containing drawing-related keywords to activate the Hunyuan text-to-image generation feature. Unleash your creativity and create any picture you desire, all for free!
You can use simple prompts similar to natural language text
画一只穿着西装的猪
draw a pig in a suit
生成一幅画，赛博朋克风，跑车
generate a painting, cyberpunk style, sports car
or multi-turn language interactions to create the picture.
画一个木制的鸟
draw a wooden bird
变成玻璃的
turn into glass
- Hunyuan-DiT (Text-to-Image Model)
- Inference
- Checkpoints
- Distillation Version
- TensorRT Version
- Training
- Lora
- Controlnet (Pose, Canny, Depth)
- IP-adapter
- Hunyuan-DiT-S checkpoints (0.7B model)
- Caption model (Re-caption the raw image-text pairs)
- DialogGen (Prompt Enhancement Model)
- Inference
- Web Demo (Gradio)
- Multi-turn T2I Demo (Gradio)
- Cli Demo
- ComfyUI
- Diffusers
- Kohya
- WebUI
- Hunyuan-DiT
We present Hunyuan-DiT, a text-to-image diffusion transformer with fine-grained understanding of both English and Chinese. To construct Hunyuan-DiT, we carefully designed the transformer structure, text encoder, and positional encoding. We also build from scratch a whole data pipeline to update and evaluate data for iterative model optimization. For fine-grained language understanding, we train a Multimodal Large Language Model to refine the captions of the images. Finally, Hunyuan-DiT can perform multi-round multi-modal dialogue with users, generating and refining images according to the context. Through our carefully designed holistic human evaluation protocol with more than 50 professional human evaluators, Hunyuan-DiT sets a new state-of-the-art in Chinese-to-image generation compared with other open-source models.
Hunyuan-DiT is a diffusion model in the latent space, as depicted in figure below. Following the Latent Diffusion Model, we use a pre-trained Variational Autoencoder (VAE) to compress the images into low-dimensional latent spaces and train a diffusion model to learn the data distribution with diffusion models. Our diffusion model is parameterized with a transformer. To encode the text prompts, we leverage a combination of pre-trained bilingual (English and Chinese) CLIP and multilingual T5 encoder.
Understanding natural language instructions and performing multi-turn interaction with users are important for a text-to-image system. It can help build a dynamic and iterative creation process that bring the user’s idea into reality step by step. In this section, we will detail how we empower Hunyuan-DiT with the ability to perform multi-round conversations and image generation. We train MLLM to understand the multi-round user dialogue and output the new text prompt for image generation.
In order to comprehensively compare the generation capabilities of HunyuanDiT and other models, we constructed a 4-dimensional test set, including Text-Image Consistency, Excluding AI Artifacts, Subject Clarity, Aesthetic. More than 50 professional evaluators performs the evaluation.
| Model | Open Source | Text-Image Consistency (%) | Excluding AI Artifacts (%) | Subject Clarity (%) | Aesthetics (%) | Overall (%) |
|---|---|---|---|---|---|---|
| SDXL | ✔ | 64.3 | 60.6 | 91.1 | 76.3 | 42.7 |
| PixArt-α | ✔ | 68.3 | 60.9 | 93.2 | 77.5 | 45.5 |
| Playground 2.5 | ✔ | 71.9 | 70.8 | 94.9 | 83.3 | 54.3 |
| SD 3 | ✘ | 77.1 | 69.3 | 94.6 | 82.5 | 56.7 |
| MidJourney v6 | ✘ | 73.5 | 80.2 | 93.5 | 87.2 | 63.3 |
| DALL-E 3 | ✘ | 83.9 | 80.3 | 96.5 | 89.4 | 71.0 |
| Hunyuan-DiT | ✔ | 74.2 | 74.3 | 95.4 | 86.6 | 59.0 |
- Chinese Elements
- Long Text Input
- Multi-turn Text2Image Generation
Hunyuan_MultiTurn_T2I_Demo.mp4
This repo consists of DialogGen (a prompt enhancement model) and Hunyuan-DiT (a text-to-image model).
The following table shows the requirements for running the models (batch size = 1):
| Model | --load-4bit (DialogGen) | GPU Peak Memory | GPU |
|---|---|---|---|
| DialogGen + Hunyuan-DiT | ✘ | 32G | A100 |
| DialogGen + Hunyuan-DiT | ✔ | 22G | A100 |
| Hunyuan-DiT | - | 11G | A100 |
| Hunyuan-DiT | - | 14G | RTX3090/RTX4090 |
- An NVIDIA GPU with CUDA support is required.
- We have tested V100 and A100 GPUs.
- Minimum: The minimum GPU memory required is 11GB.
- Recommended: We recommend using a GPU with 32GB of memory for better generation quality.
- Tested operating system: Linux
Begin by cloning the repository:
git clone https://github.com/tencent/HunyuanDiT
cd HunyuanDiTWe provide an environment.yml file for setting up a Conda environment.
Conda's installation instructions are available here.
# 1. Prepare conda environment
conda env create -f environment.yml
# 2. Activate the environment
conda activate HunyuanDiT
# 3. Install pip dependencies
python -m pip install -r requirements.txt
# 4. (Optional) Install flash attention v2 for acceleration (requires CUDA 11.6 or above)
python -m pip install git+https://github.com/Dao-AILab/flash-attention.git@v2.1.2.post3To download the model, first install the huggingface-cli. (Detailed instructions are available here.)
python -m pip install "huggingface_hub[cli]"Then download the model using the following commands:
# Create a directory named 'ckpts' where the model will be saved, fulfilling the prerequisites for running the demo.
mkdir ckpts
# Use the huggingface-cli tool to download the model.
# The download time may vary from 10 minutes to 1 hour depending on network conditions.
huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./ckpts💡Tips for using huggingface-cli (network problem)
If you encounter slow download speeds in China, you can try a mirror to speed up the download process. For example,
HF_ENDPOINT=https://hf-mirror.com huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./ckptshuggingface-cli supports resuming downloads. If the download is interrupted, you can just rerun the download
command to resume the download process.
Note: If an No such file or directory: 'ckpts/.huggingface/.gitignore.lock' like error occurs during the download
process, you can ignore the error and rerun the download command.
All models will be automatically downloaded. For more information about the model, visit the Hugging Face repository here.
| Model | #Params | Huggingface Download URL | Tencent Cloud Download URL |
|---|---|---|---|
| mT5 | 1.6B | mT5 | mT5 |
| CLIP | 350M | CLIP | CLIP |
| Tokenizer | - | Tokenizer | Tokenizer |
| DialogGen | 7.0B | DialogGen | DialogGen |
| sdxl-vae-fp16-fix | 83M | sdxl-vae-fp16-fix | sdxl-vae-fp16-fix |
| Hunyuan-DiT-v1.0 | 1.5B | Hunyuan-DiT | Hunyuan-DiT-v1.0 |
| Hunyuan-DiT-v1.1 | 1.5B | Hunyuan-DiT-v1.1 | Hunyuan-DiT-v1.1 |
| Data demo | - | - | Data demo |
Refer to the commands below to prepare the training data.
-
Install dependencies
We offer an efficient data management library, named IndexKits, supporting the management of reading hundreds of millions of data during training, see more in docs.
# 1 Install dependencies cd HunyuanDiT pip install -e ./IndexKits
-
Data download
Feel free to download the data demo.
# 2 Data download wget -O ./dataset/data_demo.zip https://dit.hunyuan.tencent.com/download/HunyuanDiT/data_demo.zip unzip ./dataset/data_demo.zip -d ./dataset mkdir ./dataset/porcelain/arrows ./dataset/porcelain/jsons -
Data conversion
Create a CSV file for training data with the fields listed in the table below.
Fields Required Description Example image_pathRequired image path ./dataset/porcelain/images/0.pngtext_zhRequired text 青花瓷风格，一只蓝色的鸟儿站在蓝色的花瓶上，周围点缀着白色花朵，背景是白色 md5Optional image md5 (Message Digest Algorithm 5) d41d8cd98f00b204e9800998ecf8427ewidthOptional image width 1024heightOptional image height 1024⚠️ Optional fields like MD5, width, and height can be omitted. If omitted, the script below will automatically calculate them. This process can be time-consuming when dealing with large-scale training data.We utilize Arrow for training data format, offering a standard and efficient in-memory data representation. A conversion script is provided to transform CSV files into Arrow format.
# 3 Data conversion python ./hydit/data_loader/csv2arrow.py ./dataset/porcelain/csvfile/image_text.csv ./dataset/porcelain/arrows -
Data Selection and Configuration File Creation
We configure the training data through YAML files. In these files, you can set up standard data processing strategies for filtering, copying, deduplicating, and more regarding the training data. For more details, see docs.
For a sample file, please refer to file. For a full parameter configuration file, see file.
-
Create training data index file using YAML file.
# Single Resolution Data Preparation idk base -c dataset/yamls/porcelain.yaml -t dataset/porcelain/jsons/porcelain.json # Multi Resolution Data Preparation idk multireso -c dataset/yamls/porcelain_mt.yaml -t dataset/porcelain/jsons/porcelain_mt.json
The directory structure for porcelain dataset is:
cd ./dataset
porcelain
├──images/ (image files)
│ ├──0.png
│ ├──1.png
│ ├──......
├──csvfile/ (csv files containing text-image pairs)
│ ├──image_text.csv
├──arrows/ (arrow files containing all necessary training data)
│ ├──00000.arrow
│ ├──00001.arrow
│ ├──......
├──jsons/ (final training data index files which read data from arrow files during training)
│ ├──porcelain.json
│ ├──porcelain_mt.jsonTo leverage DeepSpeed in training, you have the flexibility to control single-node / multi-node training by adjusting parameters such as --hostfile and --master_addr. For more details, see link.
# Single Resolution Training
PYTHONPATH=./ sh hydit/train.sh --index-file dataset/porcelain/jsons/porcelain.json
# Multi Resolution Training
PYTHONPATH=./ sh hydit/train.sh --index-file dataset/porcelain/jsons/porcelain_mt.json --multireso --reso-step 64We provide training and inference scripts for LoRA, detailed in the guidances.
# Training for porcelain LoRA.
PYTHONPATH=./ sh lora/train_lora.sh --index-file dataset/porcelain/jsons/porcelain.json
# Inference using trained LORA weights.
python sample_t2i.py --prompt "青花瓷风格，一只小狗" --no-enhance --lora-ckpt log_EXP/001-lora_porcelain_ema_rank64/checkpoints/0001000.ptWe offer two types of trained LoRA weights for porcelain and jade, see details at links
cd HunyuanDiT
# Use the huggingface-cli tool to download the model.
huggingface-cli download Tencent-Hunyuan/HYDiT-LoRA --local-dir ./ckpts/t2i/lora
# Quick start
python sample_t2i.py --prompt "青花瓷风格，一只猫在追蝴蝶" --no-enhance --load-key ema --lora-ckpt ./ckpts/t2i/lora/porcelain| Examples of training data | |||
 |
 |
 |
 |
| 青花瓷风格，一只蓝色的鸟儿站在蓝色的花瓶上，周围点缀着白色花朵，背景是白色 （Porcelain style, a blue bird stands on a blue vase, surrounded by white flowers, with a white background. ） | 青花瓷风格，这是一幅蓝白相间的陶瓷盘子，上面描绘着一只狐狸和它的幼崽在森林中漫步，背景是白色 （Porcelain style, this is a blue and white ceramic plate depicting a fox and its cubs strolling in the forest, with a white background.） | 青花瓷风格，在黑色背景上，一只蓝色的狼站在蓝白相间的盘子上，周围是树木和月亮 （Porcelain style, on a black background, a blue wolf stands on a blue and white plate, surrounded by trees and the moon.） | 青花瓷风格，在蓝色背景上，一只蓝色蝴蝶和白色花朵被放置在中央 （Porcelain style, on a blue background, a blue butterfly and white flowers are placed in the center.） |
| Examples of inference results | |||
 |
 |
 |
 |
| 青花瓷风格，苏州园林 （Porcelain style, Suzhou Gardens.） | 青花瓷风格，一朵荷花 （Porcelain style, a lotus flower.） | 青花瓷风格，一只羊（Porcelain style, a sheep.） | 青花瓷风格，一个女孩在雨中跳舞（Porcelain style, a girl dancing in the rain.） |
Make sure the conda environment is activated before running the following command.
# By default, we start a Chinese UI.
python app/hydit_app.py
# Using Flash Attention for acceleration.
python app/hydit_app.py --infer-mode fa
# You can disable the enhancement model if the GPU memory is insufficient.
# The enhancement will be unavailable until you restart the app without the `--no-enhance` flag.
python app/hydit_app.py --no-enhance
# Start with English UI
python app/hydit_app.py --lang en
# Start a multi-turn T2I generation UI.
# If your GPU memory is less than 32GB, use '--load-4bit' to enable 4-bit quantization, which requires at least 22GB of memory.
python app/multiTurnT2I_app.pyThen the demo can be accessed through http://0.0.0.0:443. It should be noted that the 0.0.0.0 here needs to be X.X.X.X with your server IP.
Please install PyTorch version 2.0 or higher in advance to satisfy the requirements of the specified version of the diffusers library.
Install 🤗 diffusers, ensuring that the version is at least 0.28.1:
pip install git+https://github.com/huggingface/diffusers.gitor
pip install diffusersYou can generate images with both Chinese and English prompts using the following Python script:
import torch
from diffusers import HunyuanDiTPipeline
pipe = HunyuanDiTPipeline.from_pretrained("Tencent-Hunyuan/HunyuanDiT-Diffusers", torch_dtype=torch.float16)
pipe.to("cuda")
# You may also use English prompt as HunyuanDiT supports both English and Chinese
# prompt = "An astronaut riding a horse"
prompt = "一个宇航员在骑马"
image = pipe(prompt).images[0]You can use our distilled model to generate images even faster:
import torch
from diffusers import HunyuanDiTPipeline
pipe = HunyuanDiTPipeline.from_pretrained("Tencent-Hunyuan/HunyuanDiT-Diffusers-Distilled", torch_dtype=torch.float16)
pipe.to("cuda")
# You may also use English prompt as HunyuanDiT supports both English and Chinese
# prompt = "An astronaut riding a horse"
prompt = "一个宇航员在骑马"
image = pipe(prompt, num_inference_steps=25).images[0]More details can be found in HunyuanDiT-Diffusers-Distilled
We provide several commands to quick start:
# Prompt Enhancement + Text-to-Image. Torch mode
python sample_t2i.py --prompt "渔舟唱晚"
# Only Text-to-Image. Torch mode
python sample_t2i.py --prompt "渔舟唱晚" --no-enhance
# Only Text-to-Image. Flash Attention mode
python sample_t2i.py --infer-mode fa --prompt "渔舟唱晚"
# Generate an image with other image sizes.
python sample_t2i.py --prompt "渔舟唱晚" --image-size 1280 768
# Prompt Enhancement + Text-to-Image. DialogGen loads with 4-bit quantization, but it may loss performance.
python sample_t2i.py --prompt "渔舟唱晚" --load-4bit
More example prompts can be found in example_prompts.txt
We list some more useful configurations for easy usage:
| Argument | Default | Description |
|---|---|---|
--prompt |
None | The text prompt for image generation |
--image-size |
1024 1024 | The size of the generated image |
--seed |
42 | The random seed for generating images |
--infer-steps |
100 | The number of steps for sampling |
--negative |
- | The negative prompt for image generation |
--infer-mode |
torch | The inference mode (torch, fa, or trt) |
--sampler |
ddpm | The diffusion sampler (ddpm, ddim, or dpmms) |
--no-enhance |
False | Disable the prompt enhancement model |
--model-root |
ckpts | The root directory of the model checkpoints |
--load-key |
ema | Load the student model or EMA model (ema or module) |
--load-4bit |
Fasle | Load DialogGen model with 4bit quantization |
We provide several commands to quick start:
# Download comfyui code
git clone https://github.com/comfyanonymous/ComfyUI.git
# Install torch, torchvision, torchaudio
pip install torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu117
# Install Comfyui essential python package
cd ComfyUI
pip install -r requirements.txt
# ComfyUI has been successfully installed!
# Download model weight as before or link the existing model folder to ComfyUI.
python -m pip install "huggingface_hub[cli]"
mkdir models/hunyuan
huggingface-cli download Tencent-Hunyuan/HunyuanDiT --local-dir ./models/hunyuan/ckpts
# Move to the ComfyUI custom_nodes folder and copy comfyui-hydit folder from HunyuanDiT Repo.
cd custom_nodes
cp -r ${HunyuanDiT}/comfyui-hydit ./
cd comfyui-hydit
# Install some essential python Package.
pip install -r requirements.txt
# Our tool has been successfully installed!
# Go to ComfyUI main folder
cd ../..
# Run the ComfyUI Lauch command
python main.py --listen --port 80
# Running ComfyUI successfully!More details can be found in ComfyUI README
We provide training scripts for ControlNet, detailed in the guidances.
# Training for canny ControlNet.
PYTHONPATH=./ sh hydit/train_controlnet.shWe offer three types of trained ControlNet weights for canny depth and pose, see details at links
cd HunyuanDiT
# Use the huggingface-cli tool to download the model.
# We recommend using distilled weights as the base model for ControlNet inference, as our provided pretrained weights are trained on them.
huggingface-cli download Tencent-Hunyuan/HYDiT-ControlNet --local-dir ./ckpts/t2i/controlnet
huggingface-cli download Tencent-Hunyuan/Distillation-v1.1 ./pytorch_model_distill.pt --local-dir ./ckpts/t2i/model
# Quick start
python3 sample_controlnet.py --no-enhance --load-key distill --infer-steps 50 --control-type canny --prompt "在夜晚的酒店门前，一座古老的中国风格的狮子雕像矗立着，它的眼睛闪烁着光芒，仿佛在守护着这座建筑。背景是夜晚的酒店前，构图方式是特写，平视，居中构图。这张照片呈现了真实摄影风格，蕴含了中国雕塑文化，同时展现了神秘氛围" --condition-image-path controlnet/asset/input/canny.jpg --control-weight 1.0| Condition Input | ||
| Canny ControlNet | Depth ControlNet | Pose ControlNet |
| 在夜晚的酒店门前，一座古老的中国风格的狮子雕像矗立着，它的眼睛闪烁着光芒，仿佛在守护着这座建筑。背景是夜晚的酒店前，构图方式是特写，平视，居中构图。这张照片呈现了真实摄影风格，蕴含了中国雕塑文化，同时展现了神秘氛围 （At night, an ancient Chinese-style lion statue stands in front of the hotel, its eyes gleaming as if guarding the building. The background is the hotel entrance at night, with a close-up, eye-level, and centered composition. This photo presents a realistic photographic style, embodies Chinese sculpture culture, and reveals a mysterious atmosphere.） |
在茂密的森林中，一只黑白相间的熊猫静静地坐在绿树红花中，周围是山川和海洋。背景是白天的森林，光线充足 （In the dense forest, a black and white panda sits quietly in green trees and red flowers, surrounded by mountains, rivers, and the ocean. The background is the forest in a bright environment.） |
一位亚洲女性，身穿绿色上衣，戴着紫色头巾和紫色围巾，站在黑板前。背景是黑板。照片采用近景、平视和居中构图的方式呈现真实摄影风格 （An Asian woman, dressed in a green top, wearing a purple headscarf and a purple scarf, stands in front of a blackboard. The background is the blackboard. The photo is presented in a close-up, eye-level, and centered composition, adopting a realistic photographic style） |
 |
 |
 |
| ControlNet Output | ||
 |
 |
 |
-
We provide TensorRT version of HunyuanDiT for inference acceleration (faster than flash attention). See Tencent-Hunyuan/TensorRT-libs for more details.
-
We provide Distillation version of HunyuanDiT for inference acceleration. See Tencent-Hunyuan/Distillation for more details.
If you find Hunyuan-DiT or DialogGen useful for your research and applications, please cite using this BibTeX:
@misc{li2024hunyuandit,
title={Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding},
author={Zhimin Li and Jianwei Zhang and Qin Lin and Jiangfeng Xiong and Yanxin Long and Xinchi Deng and Yingfang Zhang and Xingchao Liu and Minbin Huang and Zedong Xiao and Dayou Chen and Jiajun He and Jiahao Li and Wenyue Li and Chen Zhang and Rongwei Quan and Jianxiang Lu and Jiabin Huang and Xiaoyan Yuan and Xiaoxiao Zheng and Yixuan Li and Jihong Zhang and Chao Zhang and Meng Chen and Jie Liu and Zheng Fang and Weiyan Wang and Jinbao Xue and Yangyu Tao and Jianchen Zhu and Kai Liu and Sihuan Lin and Yifu Sun and Yun Li and Dongdong Wang and Mingtao Chen and Zhichao Hu and Xiao Xiao and Yan Chen and Yuhong Liu and Wei Liu and Di Wang and Yong Yang and Jie Jiang and Qinglin Lu},
year={2024},
eprint={2405.08748},
archivePrefix={arXiv},
primaryClass={cs.CV}
}
@article{huang2024dialoggen,
title={DialogGen: Multi-modal Interactive Dialogue System for Multi-turn Text-to-Image Generation},
author={Huang, Minbin and Long, Yanxin and Deng, Xinchi and Chu, Ruihang and Xiong, Jiangfeng and Liang, Xiaodan and Cheng, Hong and Lu, Qinglin and Liu, Wei},
journal={arXiv preprint arXiv:2403.08857},
year={2024}
}