Skip to content
GitLab
Menu
Projects
Groups
Snippets
Loading...
Help
Help
Support
Community forum
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
Menu
Open sidebar
Bw-bestperf
Fourcastnet_train
Commits
5f36153c
"vscode:/vscode.git/clone" did not exist on "901e7a33fa4a854957c3e19034ff5a5e2317e5f1"
Commit
5f36153c
authored
Feb 04, 2026
by
zhangwenbo
🎨
Browse files
Update README.md
parent
bdd87fae
Changes
1
Show whitespace changes
Inline
Side-by-side
Showing
1 changed file
with
32 additions
and
127 deletions
+32
-127
README.md
README.md
+32
-127
No files found.
README.md
View file @
5f36153c
...
...
@@ -11,30 +11,29 @@
### 1. 拉取镜像
```
bash
docker pull
harbor
.sourcefind.cn:5
443
/dcu/admin/base/
pytorch:2.5.1
-ubuntu22.04-dtk25.04.
4-1230
-py3.10-202
60115
docker pull
image
.sourcefind.cn:5
000
/dcu/admin/base/
vllm:0.9.2
-ubuntu22.04-dtk25.04.
1-rc5-rocblas101839-0811-das1.6
-py3.10-202
50913-rc1
```
### 2. 创建容器
```
bash
docker run -it \
--network=host \
--hostname=localhost \
--name=hunyuan \
-v /opt/hyhal:/opt/hyhal:ro \
-v $PWD:/workspace \
--ipc=host \
--device=/dev/kfd \
--device=/dev/mkfd \
--device=/dev/dri \
--shm-size=512G \
--privileged \
--group-add video \
--cap-add=SYS_PTRACE \
-u root \
--security-opt seccomp=unconfined \
harbor.sourcefind.cn:5443/dcu/admin/base/pytorch:2.5.1-ubuntu22.04-dtk25.04.4-1230-py3.10-20260115 \
/bin/bash
--network=host \
--hostname=localhost \
--name FourCastNet_train \
-v /opt/hyhal:/opt/hyhal:ro \
-v $PWD:/workspace \
--ipc=host \
--device=/dev/kfd \
--device=/dev/mkfd \
--device=/dev/dri \
--shm-size=512G \
--privileged \
--group-add video \
--cap-add=SYS_PTRACE \
--security-opt seccomp=unconfined \
image.sourcefind.cn:5000/dcu/admin/base/vllm:0.9.2-ubuntu22.04-dtk25.04.1-rc5-rocblas101839-0811-das1.6-py3.10-20250913-rc1 \
/bin/bash
```
---
...
...
@@ -44,129 +43,35 @@ harbor.sourcefind.cn:5443/dcu/admin/base/pytorch:2.5.1-ubuntu22.04-dtk25.04.4-12
### 1. 拉取代码
```
bash
git clone http://developer.sourcefind.cn/codes/bw
-
bestperf/
hunyuanvideo-i2v
.git
cd
hunyuanvideo-i2v
/
git clone http://developer.sourcefind.cn/codes/bw
_
bestperf/
fourcastnet_train
.git
cd
fourcastnet_train
/
```
### 2. 安装依赖
```
bash
pip
install
-r
requirements.txt
-i
https://pypi.tuna.tsinghua.edu.cn/simple
pip
install
yunchang
==
0.6.0
xfuser
==
0.4.2
bash fix.sh
# 适配xfuser
pip
install
wandb pybind11 h5py ruamel.yaml timm
-i
https://pypi.tuna.tsinghua.edu.cn/simple
```
### 3. 下载模型
安装 ModelScope:
```
bash
pip
install
modelscope
```
下载所需模型:
```
bash
mkdir
ckpts
modelscope download
--model
Tencent-Hunyuan/HunyuanVideo
--local_dir
./ckpts
modelscope download
--model
AI-ModelScope/HunyuanVideo-I2V
--local_dir
./ckpts
modelscope download
--model
AI-ModelScope/clip-vit-large-patch14
--local_dir
ckpts/text_encoder_2
modelscope download
--model
AI-ModelScope/llava-llama-3-8b-v1_1-transformers
--local_dir
ckpts/text_encoder_i2v
```
执行修复脚本:
```
bash
bash modified/fix.sh
```
设置环境变量,禁用 HIP 缓存 allocator 防止 OOM:
```
bash
export
PYTORCH_NO_HIP_MEMORY_CACHING
=
1
#下载数据集data 包含train和valid数据集
curl
-f
-C
-
-o
data.tar.gz https://ksefile.hpccube.com:65241/efile/s/d/ZGN1X2Vjbw
==
/fe9a61dbaca9de37
#下载的文件应按以下层次结构组织
data/
├── global_means.npy
├── global_stds.npy
├── time_means.npy
├── train/
└── valid/
```
---
## 测试代码示例(四卡多GPU测试)
导出设备环境和禁用缓存:
```
bash
export
HIP_VISIBLE_DEVICES
=
4,5,6,7
export
PYTORCH_NO_HIP_MEMORY_CACHING
=
1
```
运行文本到视频多GPU推理:
## 测试命令
```
bash
ALLOW_RESIZE_FOR_SP
=
1 torchrun
--nproc_per_node
=
4
\
sample_image2video.py
\
--model
HYVideo-T/2
\
--prompt
"An Asian man with short hair in black tactical uniform and white clothes waves a firework stick."
\
--i2v-mode
\
--i2v-image-path
./assets/demo/i2v/imgs/0.jpg
\
--i2v-resolution
720p
\
--i2v-stability
\
--infer-steps
50
\
--video-length
129
\
--flow-reverse
\
--flow-shift
17.0
\
--seed
42
\
--embedded-cfg-scale
6.0
\
--save-path
./results
\
--ulysses-degree
4
\
--ring-degree
1
\
--num-videos
1 2>&1 |
tee
z_logs/bw_image2video_4ka.log
bash run.sh
```
---
## 配置选项说明
| 参数 | 说明 | 默认值 / 示例 |
| -------------------- | ------------------------------ | ---------------------------------------- |
|
`--model`
| 指定使用的模型名称 |
`HYVideo-T/2`
|
|
`--prompt`
| 文本描述,用于生成视频 |
`"An Asian man with short hair..."`
|
|
`--i2v-mode`
| 启用文本到视频模式 | |
|
`--i2v-image-path`
| 输入图像路径 |
`./assets/demo/i2v/imgs/0.jpg`
|
|
`--i2v-resolution`
| 输出视频分辨率 |
`720p`
|
|
`--i2v-stability`
| 稳定性增强选项 | |
|
`--infer-steps`
| 推理步数,影响生成质量与速度 |
`50`
|
|
`--video-length`
| 生成视频的长度(帧数) |
`129`
|
|
`--flow-reverse`
| 是否反转光流 | |
|
`--flow-shift`
| 光流偏移值 |
`17.0`
|
|
`--seed`
| 随机种子,保证结果可复现 |
`42`
|
|
`--embedded-cfg-scale`
| Condition scaling比例 |
`6.0`
|
|
`--save-path`
| 生成结果保存路径 |
`./results`
|
|
`--ulysses-degree`
| 自定义参数,具体含义见代码说明 |
`4`
|
|
`--ring-degree`
| 自定义参数,具体含义见代码说明 |
`1`
|
|
`--num-videos`
| 生成视频数量 |
`1`
|
---
## 贡献指南
欢迎对 hunyuan-I2V 项目进行贡献!请遵循以下步骤:
1.
Fork 本仓库,并新建分支进行功能开发或问题修复。
2.
提交规范的 commit 信息,描述清晰。
3.
提交 Pull Request,简述修改内容及目的。
4.
遵守项目代码规范和测试标准。
5.
参与代码评审,积极沟通改进方案。
---
## 许可证
本项目遵循 MIT 许可证,详见
[
LICENSE
](
./LICENSE
)
文件。
---
感谢您的关注与支持!如有问题,欢迎提交 Issue 或联系维护团队。
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
.
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment