KawasakiAkasei/Hunyuan3D_2.1_Low_VRAM
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

update readme

Browse Source
This commit is contained in:
Huiwenshi
2025-06-26 16:34:51 +08:00
parent 7c92655a0d
commit e59169a8ec
4 changed files with 16 additions and 53 deletions
Download Patch File Download Diff File Expand all files Collapse all files

3
README.md
View File

@@ -33,6 +33,9 @@
|--------------------------------------------------|-------------------------------------------------------|---------------------------------------------|---------------------------------------------------|
| <img src="assets/qrcode/wechat.png" height=140> | <img src="assets/qrcode/xiaohongshu.png" height=140> | <img src="assets/qrcode/x.png" height=140> | <img src="assets/qrcode/discord.png" height=140> |
## 🤗 Community Contribution Leaderboard
1. By [@visualbruno](https://github.com/visualbruno)
- ComfyUI-Hunyuan3d-2-1: https://github.com/visualbruno/ComfyUI-Hunyuan3d-2-1
## ☯️ **Hunyuan3D 2.1**

47
hy3dshape/README-zh.md
View File

@@ -1,47 +0,0 @@
# Hunyuan3D-2.1-Shape
# 训练
我们会展示小数据集上DiT的训练全流程
## 数据预处理
渲染和水密化参考[链接](tools/README.md)，最终得到如下结构
``` yaml
dataset/preprocessed/{uid}
├── geo_data
│ ├── {uid}_sdf.npz
│ ├── {uid}_surface.npz
│ └── {uid}_watertight.obj
└── render_cond
├── 000.png
├── ...
├── 023.png
├── mesh.ply
└── transforms.json
```
我们提供了一个8个case(均来自Objaverse-XL)预处理后的结果在 tools/mini_trainset，可以直接用于过拟合训练
## 启动训练
我们提供了可供参考的训练配置文件和启动脚本（默认单机8卡deepspeed训练），用户根据需要自行修改。
配置文件
```
configs/dit-from-scratch-overfitting-flowmatching-dinog518-bf16-lr1e4-1024.yaml
```
启动脚本
```
export node_num=1
export node_rank=0
export master_ip=0.0.0.0 # set your master_ip
export config='configs/dit-from-scratch-overfitting-flowmatching-dinog518-bf16-lr1e4-1024.yaml'
export output_dir='output_folder/dit/overfitting'
bash scripts/train_deepspeed.sh $node_num $node_rank $master_ip $config $output_dir
```

15
hy3dshape/README.md
View File

@@ -45,10 +45,17 @@ configs/hunyuandit-mini-overfitting-flowmatching-dinog518-bf16-lr1e4-512.yaml
Launch Script
```
export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
export num_gpu_per_node=8
export node_num=1
export node_rank=0
export master_ip=0.0.0.0 # set your master_ip
export config=configs/hunyuandit-mini-overfitting-flowmatching-dinog518-bf16-lr1e4-512.yaml
export output_dir=output_folder/dit/overfitting
bash scripts/train_deepspeed.sh $node_num $node_rank $master_ip $config $output_dir
```
# export config=configs/hunyuandit-finetuning-flowmatching-dinol518-bf16-lr1e5-4096.yaml
# export output_dir=output_folder/dit/fintuning_lr1e5
export config=configs/hunyuandit-mini-overfitting-flowmatching-dinol518-bf16-lr1e4-4096.yaml
export output_dir=output_folder/dit/overfitting_depth_16_token_4096_lr1e4
bash scripts/train_deepspeed.sh $node_num $node_rank $num_gpu_per_node $master_ip $config $output_dir
```

4
hy3dshape/configs/hunyuandit-mini-overfitting-flowmatching-dinol518-bf16-lr1e4-4096.yaml
View File

@@ -1,6 +1,6 @@
name: "HunyuanDiT flowmatching; VAE: 4096 token length; ImageEncoder: DINO-v2 Large; ImageSize: 518"
# oversitting successfully cost 68G memory under current settings
# you can adjust model arch or batch_size according to your GPU memory
# Overfitting successfully cost 68G memory under current settings
# You can adjust model arch or batch_size according to your GPU memory
training:
steps: 10_0000_0000