From 056fe1b8933e92861bff3992945ab5a6c433e1f0 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 19:58:55 +0800
Subject: [PATCH 01/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 123 +++++++++++++++++++++
 1 file changed, 123 insertions(+)
 create mode 100644 MindIE/MultiModal/StepVideo-TI2V/README.md

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
new file mode 100644
index 0000000000..030fc25086
--- /dev/null
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -0,0 +1,123 @@
+## 一、准备运行环境
+
+  **表 1**  版本配套表
+
+  | 配套  | 版本 | 环境准备指导 |
+  | ----- | ----- |-----|
+  | Python | 3.10.2 | - |
+  | torch | 2.1.0 | - |
+
+### 1.1 获取CANN&MindIE安装包&环境准备
+- 设备支持
+Atlas 800I A2(8*64G)推理设备：支持的卡数最小为1
+- [Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
+- [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
+
+### 1.2 CANN安装
+```shell
+# 增加软件包可执行权限，{version}表示软件版本号，{arch}表示CPU架构，{soc}表示昇腾AI处理器的版本。
+chmod +x ./Ascend-cann-toolkit_{version}_linux-{arch}.run
+chmod +x ./Ascend-cann-kernels-{soc}_{version}_linux.run
+# 校验软件包安装文件的一致性和完整性
+./Ascend-cann-toolkit_{version}_linux-{arch}.run --check
+./Ascend-cann-kernels-{soc}_{version}_linux.run --check
+# 安装
+./Ascend-cann-toolkit_{version}_linux-{arch}.run --install
+./Ascend-cann-kernels-{soc}_{version}_linux.run --install
+
+# 设置环境变量
+source /usr/local/Ascend/ascend-toolkit/set_env.sh
+```
+
+### 1.3 环境依赖安装
+```shell
+pip3 install -r requirements.txt
+```
+
+### 1.4 MindIE安装
+```shell
+# 增加软件包可执行权限，{version}表示软件版本号，{arch}表示CPU架构。
+chmod +x ./Ascend-mindie_${version}_linux-${arch}.run
+./Ascend-mindie_${version}_linux-${arch}.run --check
+
+# 方式一：默认路径安装
+./Ascend-mindie_${version}_linux-${arch}.run --install
+# 设置环境变量
+cd /usr/local/Ascend/mindie && source set_env.sh
+
+# 方式二：指定路径安装
+./Ascend-mindie_${version}_linux-${arch}.run --install-path=${AieInstallPath}
+# 设置环境变量
+cd ${AieInstallPath}/mindie && source set_env.sh
+```
+
+### 1.5 Torch_npu安装
+下载 pytorch_v{pytorchversion}_py{pythonversion}.tar.gz
+```shell
+tar -xzvf pytorch_v{pytorchversion}_py{pythonversion}.tar.gz
+# 解压后，会有whl包
+pip install torch_npu-{pytorchversion}.xxxx.{arch}.whl
+```
+
+## 二、下载权重
+
+### 2.1 权重及配置文件说明
+stepvideo-ti2v权重链接:
+```shell
+https://huggingface.co/stepfun-ai/stepvideo-ti2v
+```
+
+## 三、StepVideo-TI2V使用
+当前支持的分辨率：
+| 分辨率 | h/w=9:16 | h/w=1:1 |
+| ---- | ---- | ---- |
+| 720P | 544x1992 | 768x768 |
+### 3.1 下载到本地
+```shell
+git clone https://modelers.cn/MindIE/stepvideo-ti2v.git
+cd stepvideo-ti2v
+```
+
+
+### 3.2 性能测试
+执行命令：
+```shell
+# 使用上一步下载的权重
+export model_dir='./stepvideo-ti2v/'
+export ALGO=0
+
+# 指定0卡和1卡开启vae和text_encoder的api服务
+ASCEND_RT_VISIBLE_DEVICES=0 python api/call_remote_server.py --model_dir ${model_dir} --enable_vae True &
+ASCEND_RT_VISIBLE_DEVICES=1 python api/call_remote_server.py --model_dir ${model_dir} --enable_llm True &
+
+# 执行推理
+ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
+--model_dir ${model_dir} \
+--height 554 \
+--width 992 \
+--num_frames 102 \
+--infer_steps 50  \
+--ulysses_degree 2 \
+--tensor_parallel_degree 2 \
+--prompt="一名宇航员在月球上发现一块石碑，上面印有“stepfun”字样，闪闪发光"  \
+--first_image_path './assets/demo.png' \
+--save_path './results'
+```
+
+参数说明：
+- ALGO: 为0表示默认FA算子；设置为1表示使用高性能FA算子
+- model_dir: 配置文件及权重路径。
+- height: 生成视频的高
+- weight: 生成视频的宽
+- num_frames: 视频帧数
+- infer_steps: 推理步数
+- ulysses_degree: ulysses并行度
+- tensor_parallel_degree: tp并行度
+- prompt: 文本提示词
+- first_image_path: 用于生成视频的图片路径
+- save_path: 生成的视频的保存路径
+
+
+## 声明
+- 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
+- 如您在使用本代码仓的过程中，发现任何问题（包括但不限于功能问题、合规问题），请在本代码仓提交issue，我们将及时审视并解答。
\ No newline at end of file
-- 
Gitee


From 318c129a6d60abbf49c3233934889fceae797155 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 20:06:25 +0800
Subject: [PATCH 02/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 5 +----
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 030fc25086..6293082f7d 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -68,10 +68,7 @@ https://huggingface.co/stepfun-ai/stepvideo-ti2v
 ```
 
 ## 三、StepVideo-TI2V使用
-当前支持的分辨率：
-| 分辨率 | h/w=9:16 | h/w=1:1 |
-| ---- | ---- | ---- |
-| 720P | 544x1992 | 768x768 |
+当前支持的分辨率：544x992、768x768
 ### 3.1 下载到本地
 ```shell
 git clone https://modelers.cn/MindIE/stepvideo-ti2v.git
-- 
Gitee


From 3c133d33e8ad69caa7d817c5bcd786cf3b43079c Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 20:09:42 +0800
Subject: [PATCH 03/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 6293082f7d..4166f5dffd 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -75,7 +75,6 @@ git clone https://modelers.cn/MindIE/stepvideo-ti2v.git
 cd stepvideo-ti2v
 ```
 
-
 ### 3.2 性能测试
 执行命令：
 ```shell
@@ -114,6 +113,11 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
 
+## 四、模型推理性能结果参考
+### StepVideo-TI2V
+| 硬件形态  | cpu规格 | 卡数 | 迭代次数 | 等价优化平均耗时 | 算法优化平均耗时  |
+| :------: | :------: | :------: |:----:| :------: |:-----:|
+| Atlas 800I A2(8*64G) | 64核(x86) |  1  |  50  |  |  |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 96fb70c9fbebe2d5ea0105e40150f590ebc63358 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 20:11:01 +0800
Subject: [PATCH 04/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 4166f5dffd..799bb7c97c 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -115,9 +115,9 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 卡数 | 迭代次数 | 等价优化平均耗时 | 算法优化平均耗时  |
-| :------: | :------: | :------: |:----:| :------: |:-----:|
-| Atlas 800I A2(8*64G) | 64核(x86) |  1  |  50  |  |  |
+| 硬件形态  | cpu规格 | 卡数 | 迭代次数 | 平均耗时 |
+| :------: | :------: | :------: |:----:|:-----:|
+| Atlas 800I A2(8*64G) | 64核(x86) |  4  |  50  |  |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 299857dcfb4a50dbf8fee7562bd2f956a1ef2cf4 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 21:27:41 +0800
Subject: [PATCH 05/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 799bb7c97c..c44020765c 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -63,7 +63,13 @@ pip install torch_npu-{pytorchversion}.xxxx.{arch}.whl
 
 ### 2.1 权重及配置文件说明
 stepvideo-ti2v权重链接:
+
+```shell
+# modelers
+https://modelers.cn/StepFun/stepvideo-ti2v
+```
 ```shell
+# huggingface
 https://huggingface.co/stepfun-ai/stepvideo-ti2v
 ```
 
-- 
Gitee


From b13d6e3380566ec243b82c890cd2cd36169b3cf7 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Mon, 17 Mar 2025 21:48:06 +0800
Subject: [PATCH 06/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index c44020765c..4e99cde2fa 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -66,7 +66,7 @@ stepvideo-ti2v权重链接:
 
 ```shell
 # modelers
-https://modelers.cn/StepFun/stepvideo-ti2v
+https://modelers.cn/models/StepFun/stepvideo-ti2v
 ```
 ```shell
 # huggingface
-- 
Gitee


From 411f1870d5c5a3f6ec5462463f2f8ae545da1851 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Tue, 18 Mar 2025 14:33:19 +0800
Subject: [PATCH 07/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 4e99cde2fa..19d0dd0764 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,15 +1,15 @@
+## 版本配套
+| 组件 | 版本 |
+| - | - |
+| MindIE | 1.0.0 |
+| CANN | 8.0.0 |
+| PTA | 6.0.0 |
+| MindStudio | 7.0.0 |
+| HDK | 24.1.0 |
 ## 一、准备运行环境
-
-  **表 1**  版本配套表
-
-  | 配套  | 版本 | 环境准备指导 |
-  | ----- | ----- |-----|
-  | Python | 3.10.2 | - |
-  | torch | 2.1.0 | - |
-
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
-Atlas 800I A2(8*64G)推理设备：支持的卡数最小为1
+Atlas 800I A2(8*64G)推理设备：支持的卡数最小为6
 - [Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
 
-- 
Gitee


From 9d6c06ab9da1e9358a65bbe0345b894669138c5a Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Tue, 18 Mar 2025 15:58:05 +0800
Subject: [PATCH 08/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 10 +++-------
 1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 19d0dd0764..1f5818c169 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -9,6 +9,7 @@
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
+Atlas 800T A2(8*64G)推理设备：支持的卡数最小为6
 Atlas 800I A2(8*64G)推理设备：支持的卡数最小为6
 - [Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
@@ -99,8 +100,8 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 --width 992 \
 --num_frames 102 \
 --infer_steps 50  \
---ulysses_degree 2 \
---tensor_parallel_degree 2 \
+--ulysses_degree 4 \
+--tensor_parallel_degree 1 \
 --prompt="一名宇航员在月球上发现一块石碑，上面印有“stepfun”字样，闪闪发光"  \
 --first_image_path './assets/demo.png' \
 --save_path './results'
@@ -119,11 +120,6 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
 
-## 四、模型推理性能结果参考
-### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 卡数 | 迭代次数 | 平均耗时 |
-| :------: | :------: | :------: |:----:|:-----:|
-| Atlas 800I A2(8*64G) | 64核(x86) |  4  |  50  |  |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 53f71cde1ffc23672334ecd52ffea31230f1606d Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Tue, 18 Mar 2025 16:06:11 +0800
Subject: [PATCH 09/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 1f5818c169..72fe2b99e1 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -11,7 +11,7 @@
 - 设备支持
 Atlas 800T A2(8*64G)推理设备：支持的卡数最小为6
 Atlas 800I A2(8*64G)推理设备：支持的卡数最小为6
-- [Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
+- [Atlas 800T A2(8*64G)/Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
 
 ### 1.2 CANN安装
-- 
Gitee


From 0e85ae78194876b95a7867ca14d0bf166603e3a9 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Tue, 18 Mar 2025 16:09:35 +0800
Subject: [PATCH 10/34] add_stepvideo-ti2v

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 72fe2b99e1..51aece08ec 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -9,9 +9,9 @@
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
-Atlas 800T A2(8*64G)推理设备：支持的卡数最小为6
-Atlas 800I A2(8*64G)推理设备：支持的卡数最小为6
-- [Atlas 800T A2(8*64G)/Atlas 800I A2(8*64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
+Atlas 800T A2(64G)：支持的卡数最小为6
+Atlas 800I A2(64G)：支持的卡数最小为6
+- [Atlas 800T A2(64G)/Atlas 800I A2(64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
 
 ### 1.2 CANN安装
-- 
Gitee


From 44f8f2751c7c098cdbe0076ae3af38058b23341b Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Tue, 18 Mar 2025 16:48:59 +0800
Subject: [PATCH 11/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 51aece08ec..1acefe624d 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -78,7 +78,7 @@ https://huggingface.co/stepfun-ai/stepvideo-ti2v
 当前支持的分辨率：544x992、768x768
 ### 3.1 下载到本地
 ```shell
-git clone https://modelers.cn/MindIE/stepvideo-ti2v.git
+git clone https://modelers.cn/MindIE/Step-Video-TI2V.git
 cd stepvideo-ti2v
 ```
 
-- 
Gitee


From 310cf003c18e6dda1d41a9661e3c83c6be5e82c0 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 14:50:29 +0800
Subject: [PATCH 12/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 39 ++++++++++------------
 1 file changed, 17 insertions(+), 22 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 1acefe624d..9f6824e9f0 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,11 +1,3 @@
-## 版本配套
-| 组件 | 版本 |
-| - | - |
-| MindIE | 1.0.0 |
-| CANN | 8.0.0 |
-| PTA | 6.0.0 |
-| MindStudio | 7.0.0 |
-| HDK | 24.1.0 |
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
@@ -30,12 +22,7 @@ chmod +x ./Ascend-cann-kernels-{soc}_{version}_linux.run
 source /usr/local/Ascend/ascend-toolkit/set_env.sh
 ```
 
-### 1.3 环境依赖安装
-```shell
-pip3 install -r requirements.txt
-```
-
-### 1.4 MindIE安装
+### 1.3 MindIE安装
 ```shell
 # 增加软件包可执行权限，{version}表示软件版本号，{arch}表示CPU架构。
 chmod +x ./Ascend-mindie_${version}_linux-${arch}.run
@@ -52,7 +39,7 @@ cd /usr/local/Ascend/mindie && source set_env.sh
 cd ${AieInstallPath}/mindie && source set_env.sh
 ```
 
-### 1.5 Torch_npu安装
+### 1.4 Torch_npu安装
 下载 pytorch_v{pytorchversion}_py{pythonversion}.tar.gz
 ```shell
 tar -xzvf pytorch_v{pytorchversion}_py{pythonversion}.tar.gz
@@ -79,15 +66,24 @@ https://huggingface.co/stepfun-ai/stepvideo-ti2v
 ### 3.1 下载到本地
 ```shell
 git clone https://modelers.cn/MindIE/Step-Video-TI2V.git
-cd stepvideo-ti2v
+```
+### 3.2 安装依赖
+```shell
+cd StepVideo-TI2V
+pip install -e .
+```
+安装xfuser
+```shell
+git clone -b 0.4.2 https://github.com/xdit-project/xDiT.git
+bash patch.sh
 ```
 
-### 3.2 性能测试
+### 3.3 性能测试
 执行命令：
 ```shell
 # 使用上一步下载的权重
 export model_dir='./stepvideo-ti2v/'
-export ALGO=0
+export ALGO=1
 
 # 指定0卡和1卡开启vae和text_encoder的api服务
 ASCEND_RT_VISIBLE_DEVICES=0 python api/call_remote_server.py --model_dir ${model_dir} --enable_vae True &
@@ -100,9 +96,9 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 --width 992 \
 --num_frames 102 \
 --infer_steps 50  \
---ulysses_degree 4 \
---tensor_parallel_degree 1 \
---prompt="一名宇航员在月球上发现一块石碑，上面印有“stepfun”字样，闪闪发光"  \
+--ulysses_degree 1 \
+--tensor_parallel_degree 4 \
+--prompt="笑起来"  \
 --first_image_path './assets/demo.png' \
 --save_path './results'
 ```
@@ -120,7 +116,6 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
 
-
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
 - 如您在使用本代码仓的过程中，发现任何问题（包括但不限于功能问题、合规问题），请在本代码仓提交issue，我们将及时审视并解答。
\ No newline at end of file
-- 
Gitee


From 8a8ad58d68a702ebe6b3f5b6b1a34632cc6f2a67 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 15:02:35 +0800
Subject: [PATCH 13/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 9f6824e9f0..c1de02daa7 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,3 +1,7 @@
+## 版本配套
+敬请期待
+
+
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
@@ -116,6 +120,9 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
 
+## 四、模型推理性能结果参考
+尽情期待
+
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
 - 如您在使用本代码仓的过程中，发现任何问题（包括但不限于功能问题、合规问题），请在本代码仓提交issue，我们将及时审视并解答。
\ No newline at end of file
-- 
Gitee


From 7d724a7e86a8507e45bd6285ec0bbb9a5aa05c83 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 15:40:26 +0800
Subject: [PATCH 14/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index c1de02daa7..e46b475212 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -121,7 +121,11 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - save_path: 生成的视频的保存路径
 
 ## 四、模型推理性能结果参考
-尽情期待
+### StepVideo-TI2V
+| 硬件形态  | cpu规格 | 分辨率 | 迭代次数 | 平均耗时 | 
+| :------: | :------: | :------: |:----:| :------: |
+| G8600 | 64核(x86) | 544x992 | 50  |  |
+| G8600 | 64核(x86) | 768x768 | 50  |  |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 6cb408e09162a3dd1ee2d77c7df97caa8209d943 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 16:53:57 +0800
Subject: [PATCH 15/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index e46b475212..0439b0952a 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -124,8 +124,8 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 ### StepVideo-TI2V
 | 硬件形态  | cpu规格 | 分辨率 | 迭代次数 | 平均耗时 | 
 | :------: | :------: | :------: |:----:| :------: |
-| G8600 | 64核(x86) | 544x992 | 50  |  |
-| G8600 | 64核(x86) | 768x768 | 50  |  |
+| G8600 | 64核(x86) | 544x992 | 50  | 434s |
+| G8600 | 64核(x86) | 768x768 | 50  | 507s |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 98c758c0d653b0606d3f81880578a115a2db5998 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 17:46:29 +0800
Subject: [PATCH 16/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 36 +++++++++++++++++-----
 1 file changed, 28 insertions(+), 8 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 0439b0952a..335089a659 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -84,16 +84,19 @@ bash patch.sh
 
 ### 3.3 性能测试
 执行命令：
+使用上一步下载的权重
 ```shell
-# 使用上一步下载的权重
 export model_dir='./stepvideo-ti2v/'
 export ALGO=1
-
-# 指定0卡和1卡开启vae和text_encoder的api服务
+```
+指定0卡和1卡开启vae和text_encoder的api服务
+```shell
 ASCEND_RT_VISIBLE_DEVICES=0 python api/call_remote_server.py --model_dir ${model_dir} --enable_vae True &
 ASCEND_RT_VISIBLE_DEVICES=1 python api/call_remote_server.py --model_dir ${model_dir} --enable_llm True &
+```
 
-# 执行推理
+四卡TP并行
+```shell
 ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 --model_dir ${model_dir} \
 --height 554 \
@@ -102,6 +105,22 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 --infer_steps 50  \
 --ulysses_degree 1 \
 --tensor_parallel_degree 4 \
+--use-cpu-offload
+--prompt="笑起来"  \
+--first_image_path './assets/demo.png' \
+--save_path './results'
+```
+
+八卡TP并行
+```shell
+ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parallel.py \
+--model_dir ${model_dir} \
+--height 554 \
+--width 992 \
+--num_frames 102 \
+--infer_steps 50  \
+--ulysses_degree 1 \
+--tensor_parallel_degree 8 \
 --prompt="笑起来"  \
 --first_image_path './assets/demo.png' \
 --save_path './results'
@@ -116,16 +135,17 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 - infer_steps: 推理步数
 - ulysses_degree: ulysses并行度
 - tensor_parallel_degree: tp并行度
+- use-cpu-offload: 对DiT使用cpu-offload
 - prompt: 文本提示词
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 分辨率 | 迭代次数 | 平均耗时 | 
-| :------: | :------: | :------: |:----:| :------: |
-| G8600 | 64核(x86) | 544x992 | 50  | 434s |
-| G8600 | 64核(x86) | 768x768 | 50  | 507s |
+| 硬件形态  | cpu规格 | 分辨率 | tp并行度 | 迭代次数 | 平均耗时 | 
+| :------: | :------: | :------: | :------: |:----:| :------: |
+| G8600 | 64核(x86) | 544x992 | 8 | 50  | 434s |
+| G8600 | 64核(x86) | 768x768 | 8 | 50  | 507s |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From a74a473462e429a2bed6e49fb33be132ae5f1758 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 17:52:30 +0800
Subject: [PATCH 17/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 335089a659..dd7aafab03 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -105,7 +105,7 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 --infer_steps 50  \
 --ulysses_degree 1 \
 --tensor_parallel_degree 4 \
---use-cpu-offload
+--use_cpu_offload \
 --prompt="笑起来"  \
 --first_image_path './assets/demo.png' \
 --save_path './results'
@@ -135,7 +135,7 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parall
 - infer_steps: 推理步数
 - ulysses_degree: ulysses并行度
 - tensor_parallel_degree: tp并行度
-- use-cpu-offload: 对DiT使用cpu-offload
+- use_cpu_offload: 对DiT使用cpu-offload
 - prompt: 文本提示词
 - first_image_path: 用于生成视频的图片路径
 - save_path: 生成的视频的保存路径
-- 
Gitee


From cedf7d11affac5a39be0dc627de310b3ffa057dc Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 17:58:40 +0800
Subject: [PATCH 18/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index dd7aafab03..55e095b9dc 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,5 +1,5 @@
 ## 版本配套
-敬请期待
+### 敬请期待
 
 
 ## 一、准备运行环境
@@ -144,8 +144,8 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parall
 ### StepVideo-TI2V
 | 硬件形态  | cpu规格 | 分辨率 | tp并行度 | 迭代次数 | 平均耗时 | 
 | :------: | :------: | :------: | :------: |:----:| :------: |
-| G8600 | 64核(x86) | 544x992 | 8 | 50  | 434s |
-| G8600 | 64核(x86) | 768x768 | 8 | 50  | 507s |
+| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 50  | 434s |
+| Atlas 800T A2(64G) | 64核(x86) | 768x768 | 8 | 50  | 507s |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 3660f0123ff47f7e55438c5eefbc19b8cd8c960b Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 18:10:42 +0800
Subject: [PATCH 19/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 55e095b9dc..b364f8b19a 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -142,7 +142,7 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parall
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 分辨率 | tp并行度 | 迭代次数 | 平均耗时 | 
+| 硬件形态  | cpu规格 | 分辨率 | tp并行度 | 迭代次数 | 端到端耗时 | 
 | :------: | :------: | :------: | :------: |:----:| :------: |
 | Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 50  | 434s |
 | Atlas 800T A2(64G) | 64核(x86) | 768x768 | 8 | 50  | 507s |
-- 
Gitee


From f631d813e3a9331cfaeff0747f93613db8dd38b6 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:01:55 +0800
Subject: [PATCH 20/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index b364f8b19a..582f90ec6a 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -142,10 +142,10 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parall
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 分辨率 | tp并行度 | 迭代次数 | 端到端耗时 | 
-| :------: | :------: | :------: | :------: |:----:| :------: |
-| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 50  | 434s |
-| Atlas 800T A2(64G) | 64核(x86) | 768x768 | 8 | 50  | 507s |
+| 硬件形态  | cpu规格 | 分辨率 | 并行度 | 迭代次数 | 端到端耗时(w/o cache) | 端到端耗时(with cache) |
+| :------: | :------: | :------: | :------: |:----:| :------: | :------: |
+| A800 | 64核(x86) | 544x992 | 5 | 30  | 491s | \ |
+| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 30  | 271s | 236s |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 8fde4bb39594ffea9f14620af53a246f4e30a292 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:03:57 +0800
Subject: [PATCH 21/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 7 +------
 1 file changed, 1 insertion(+), 6 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 582f90ec6a..a004a5b43f 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -89,11 +89,6 @@ bash patch.sh
 export model_dir='./stepvideo-ti2v/'
 export ALGO=1
 ```
-指定0卡和1卡开启vae和text_encoder的api服务
-```shell
-ASCEND_RT_VISIBLE_DEVICES=0 python api/call_remote_server.py --model_dir ${model_dir} --enable_vae True &
-ASCEND_RT_VISIBLE_DEVICES=1 python api/call_remote_server.py --model_dir ${model_dir} --enable_llm True &
-```
 
 四卡TP并行
 ```shell
@@ -113,7 +108,7 @@ ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
 
 八卡TP并行
 ```shell
-ASCEND_RT_VISIBLE_DEVICES=2,3,4,5,6,7,8,9 torchrun --nproc_per_node 8 run_parallel.py \
+torchrun --nproc_per_node 8 run_parallel.py \
 --model_dir ${model_dir} \
 --height 554 \
 --width 992 \
-- 
Gitee


From 4e317b6663160e09ee9b19c078cf36c588d3bca5 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:06:09 +0800
Subject: [PATCH 22/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index a004a5b43f..98a3403a63 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -92,7 +92,7 @@ export ALGO=1
 
 四卡TP并行
 ```shell
-ASCEND_RT_VISIBLE_DEVICES=2,3,4,5 torchrun --nproc_per_node 4 run_parallel.py \
+torchrun --nproc_per_node 4 run_parallel.py \
 --model_dir ${model_dir} \
 --height 554 \
 --width 992 \
-- 
Gitee


From 821718b33f99e2dd3b49ef792c4d2d473704655c Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:09:45 +0800
Subject: [PATCH 23/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 98a3403a63..a9e0017224 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -137,10 +137,10 @@ torchrun --nproc_per_node 8 run_parallel.py \
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 分辨率 | 并行度 | 迭代次数 | 端到端耗时(w/o cache) | 端到端耗时(with cache) |
+| 硬件形态  | cpu规格 | 分辨率 | 并行度 | 迭代次数 | 端到端耗时 | 单卡吞吐 |
 | :------: | :------: | :------: | :------: |:----:| :------: | :------: |
-| A800 | 64核(x86) | 544x992 | 5 | 30  | 491s | \ |
-| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 30  | 271s | 236s |
+| A800 | 64核(x86) | 544x992 | 5 | 30  | 491s | 0.041547 |
+| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 30  | 236s | 0.053571 |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 8048d03de9f4e28bbfb4bf2a0f9502144172f48e Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:24:29 +0800
Subject: [PATCH 24/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 15 ++++++---------
 1 file changed, 6 insertions(+), 9 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index a9e0017224..344f9d9d4d 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,7 +1,3 @@
-## 版本配套
-### 敬请期待
-
-
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
@@ -9,7 +5,8 @@ Atlas 800T A2(64G)：支持的卡数最小为6
 Atlas 800I A2(64G)：支持的卡数最小为6
 - [Atlas 800T A2(64G)/Atlas 800I A2(64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
-
+- MindIE配套版本: 敬请期待
+  
 ### 1.2 CANN安装
 ```shell
 # 增加软件包可执行权限，{version}表示软件版本号，{arch}表示CPU架构，{soc}表示昇腾AI处理器的版本。
@@ -137,10 +134,10 @@ torchrun --nproc_per_node 8 run_parallel.py \
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
-| 硬件形态  | cpu规格 | 分辨率 | 并行度 | 迭代次数 | 端到端耗时 | 单卡吞吐 |
-| :------: | :------: | :------: | :------: |:----:| :------: | :------: |
-| A800 | 64核(x86) | 544x992 | 5 | 30  | 491s | 0.041547 |
-| Atlas 800T A2(64G) | 64核(x86) | 544x992 | 8 | 30  | 236s | 0.053571 |
+| 硬件形态  | 分辨率 | GPU/NPU | 迭代次数 | 端到端耗时 | 单卡吞吐 |
+| :------: | :------: | :------: |:----:| :------: | :------: |
+| A800 | 544x992 | 5 | 30  | 491s | 0.041547 |
+| Atlas 800T A2(64G)| 544x992 | 8 | 30  | 236s | 0.053571 |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From a9d022aac84db77159f79b8ae49c705b0a445773 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:26:57 +0800
Subject: [PATCH 25/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 344f9d9d4d..cf7b4edebf 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -55,7 +55,7 @@ stepvideo-ti2v权重链接:
 
 ```shell
 # modelers
-https://modelers.cn/models/StepFun/stepvideo-ti2v
+https://modelscope.cn/models/stepfun-ai/stepvideo-ti2v
 ```
 ```shell
 # huggingface
-- 
Gitee


From 376bf006f6f25f29525f2029157d4eec3c3c5a4d Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:29:14 +0800
Subject: [PATCH 26/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 13 +++----------
 1 file changed, 3 insertions(+), 10 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index cf7b4edebf..86d276b8ac 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -51,16 +51,9 @@ pip install torch_npu-{pytorchversion}.xxxx.{arch}.whl
 ## 二、下载权重
 
 ### 2.1 权重及配置文件说明
-stepvideo-ti2v权重链接:
-
-```shell
-# modelers
-https://modelscope.cn/models/stepfun-ai/stepvideo-ti2v
-```
-```shell
-# huggingface
-https://huggingface.co/stepfun-ai/stepvideo-ti2v
-```
+|     Models      |                        🤗 Huggingface                         |                         🤖 Modelscope                         |
+| :-------------: | :----------------------------------------------------------: | :----------------------------------------------------------: |
+| Step-Video-TI2V | [Download](https://huggingface.co/stepfun-ai/stepvideo-ti2v) | [Download](https://modelscope.cn/models/stepfun-ai/stepvideo-ti2v) |
 
 ## 三、StepVideo-TI2V使用
 当前支持的分辨率：544x992、768x768
-- 
Gitee


From fc587ccfdb4ab0636cea664bc90eb934e4d4a9a9 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:42:14 +0800
Subject: [PATCH 27/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 8 +++++++-
 1 file changed, 7 insertions(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 86d276b8ac..7da4a9dd58 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -1,3 +1,7 @@
+## 简述
+Step-Video-TI2V，一款参数规模达30B的SOTA文本驱动图像转视频生成模型，基于文本与图像输入，可生成长达 102 帧的视频。该模型在Step-Video-T2V基础上完成训练,将图像条件融入生成视频的首帧，借助 Step-Video-T2V 的视频变分自编码器（Video-VAE）将其编码为潜在表征。此外，引入运动分数条件，使得用户能够调控基于图像条件生成视频的动态程度。
+
+通过MindIE SD推理适配，实现了Step-Video-TI2V模型在昇腾上的高性能推理。主要使用了FA长序列场景下的高性能模式和混合并行切分策略，执行544px × 992px × 102f的视频生成（step=30），耗时236s
 ## 一、准备运行环境
 ### 1.1 获取CANN&MindIE安装包&环境准备
 - 设备支持
@@ -5,7 +9,8 @@ Atlas 800T A2(64G)：支持的卡数最小为6
 Atlas 800I A2(64G)：支持的卡数最小为6
 - [Atlas 800T A2(64G)/Atlas 800I A2(64G)](https://www.hiascend.com/developer/download/community/result?module=pt+ie+cann&product=4&model=32)
 - [环境准备指导](https://www.hiascend.com/document/detail/zh/CANNCommunityEdition/80RC2alpha002/softwareinst/instg/instg_0001.html)
-- MindIE配套版本: 敬请期待
+- MindIE配套版本
+   **敬请期待**
   
 ### 1.2 CANN安装
 ```shell
@@ -127,6 +132,7 @@ torchrun --nproc_per_node 8 run_parallel.py \
 
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
+NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
 | 硬件形态  | 分辨率 | GPU/NPU | 迭代次数 | 端到端耗时 | 单卡吞吐 |
 | :------: | :------: | :------: |:----:| :------: | :------: |
 | A800 | 544x992 | 5 | 30  | 491s | 0.041547 |
-- 
Gitee


From 7aa3b1c590bed087d605f222ecf9821ebb441186 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:43:38 +0800
Subject: [PATCH 28/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 7da4a9dd58..4f6beb9d8f 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -133,7 +133,7 @@ torchrun --nproc_per_node 8 run_parallel.py \
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
 NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
-| 硬件形态  | 分辨率 | GPU/NPU | 迭代次数 | 端到端耗时 | 单卡吞吐 |
+| 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐 |
 | :------: | :------: | :------: |:----:| :------: | :------: |
 | A800 | 544x992 | 5 | 30  | 491s | 0.041547 |
 | Atlas 800T A2(64G)| 544x992 | 8 | 30  | 236s | 0.053571 |
-- 
Gitee


From 5501e5ccfe0dada607d1279e8bb543ce462201ca Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:46:17 +0800
Subject: [PATCH 29/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 4f6beb9d8f..0836810c04 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -133,7 +133,7 @@ torchrun --nproc_per_node 8 run_parallel.py \
 ## 四、模型推理性能结果参考
 ### StepVideo-TI2V
 NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
-| 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐 |
+| 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐（fps/p） |
 | :------: | :------: | :------: |:----:| :------: | :------: |
 | A800 | 544x992 | 5 | 30  | 491s | 0.041547 |
 | Atlas 800T A2(64G)| 544x992 | 8 | 30  | 236s | 0.053571 |
-- 
Gitee


From 77dba8e6e719c5a05936413b154c13a72c46cc2f Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 21:56:12 +0800
Subject: [PATCH 30/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 0836810c04..7c39b69c02 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -135,8 +135,8 @@ torchrun --nproc_per_node 8 run_parallel.py \
 NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
 | 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐（fps/p） |
 | :------: | :------: | :------: |:----:| :------: | :------: |
-| A800 | 544x992 | 5 | 30  | 491s | 0.041547 |
-| Atlas 800T A2(64G)| 544x992 | 8 | 30  | 236s | 0.053571 |
+| A800 | 102fx544x992 | 5 | 30  | 491s | 0.041547 |
+| Atlas 800T A2(64G)| 102fx544x992 | 8 | 30  | 236s | 0.053571 |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 9197000de149df9c56869360cc36586a77e5b705 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 22:05:45 +0800
Subject: [PATCH 31/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 7c39b69c02..b027f27476 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -135,8 +135,8 @@ torchrun --nproc_per_node 8 run_parallel.py \
 NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
 | 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐（fps/p） |
 | :------: | :------: | :------: |:----:| :------: | :------: |
-| A800 | 102fx544x992 | 5 | 30  | 491s | 0.041547 |
-| Atlas 800T A2(64G)| 102fx544x992 | 8 | 30  | 236s | 0.053571 |
+| A800 | 544px x 992px x 102f | 5 | 30  | 491s | 0.041547 |
+| Atlas 800T A2(64G)| 544px x 992px x 102f | 8 | 30  | 236s | 0.053571 |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 9dc7adf9f463078498cf5768d215398929569736 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 22:09:06 +0800
Subject: [PATCH 32/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index b027f27476..9729cced5b 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -135,8 +135,8 @@ torchrun --nproc_per_node 8 run_parallel.py \
 NPU端到端性能和GPU进行了对比，平均每卡的吞吐达到1.29x GPU A800。性能测试如下：
 | 硬件形态  | 分辨率 | GPU数（NPU数） | 迭代次数 | 端到端耗时 | 单卡吞吐（fps/p） |
 | :------: | :------: | :------: |:----:| :------: | :------: |
-| A800 | 544px x 992px x 102f | 5 | 30  | 491s | 0.041547 |
-| Atlas 800T A2(64G)| 544px x 992px x 102f | 8 | 30  | 236s | 0.053571 |
+| A800 | 544px × 992px × 102f | 5 | 30  | 491s | 0.041547 |
+| Atlas 800T A2(64G)| 544px × 992px × 102f | 8 | 30  | 236s | 0.053571 |
 
 ## 声明
 - 本代码仓提到的数据集和模型仅作为示例，这些数据集和模型仅供您用于非商业目的，如您使用这些数据集和模型来完成示例，请您特别注意应遵守对应数据集和模型的License，如您因使用数据集或模型而产生侵权纠纷，华为不承担任何责任。
-- 
Gitee


From 14315ac4bf71f03125e76ba0fa44b7f5137b02a7 Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 22:18:44 +0800
Subject: [PATCH 33/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index 9729cced5b..f917851d31 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -56,9 +56,9 @@ pip install torch_npu-{pytorchversion}.xxxx.{arch}.whl
 ## 二、下载权重
 
 ### 2.1 权重及配置文件说明
-|     Models      |                        🤗 Huggingface                         |                         🤖 Modelscope                         |
+|     Models      |                        🤗 Huggingface                         |                         MODELERS                          |
 | :-------------: | :----------------------------------------------------------: | :----------------------------------------------------------: |
-| Step-Video-TI2V | [Download](https://huggingface.co/stepfun-ai/stepvideo-ti2v) | [Download](https://modelscope.cn/models/stepfun-ai/stepvideo-ti2v) |
+| Step-Video-TI2V | [Download](https://huggingface.co/stepfun-ai/stepvideo-ti2v) | [Download](https://modelers.cn/models/StepFun/Step-Video-TI2V-npu-weight) |
 
 ## 三、StepVideo-TI2V使用
 当前支持的分辨率：544x992、768x768
-- 
Gitee


From 4caa6861c8feb4d0504cc7a1a4b966876e172a2a Mon Sep 17 00:00:00 2001
From: zhoufan2956 <zhoufan2956@163.com>
Date: Wed, 19 Mar 2025 22:19:54 +0800
Subject: [PATCH 34/34] update

---
 MindIE/MultiModal/StepVideo-TI2V/README.md | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/MindIE/MultiModal/StepVideo-TI2V/README.md b/MindIE/MultiModal/StepVideo-TI2V/README.md
index f917851d31..c420c9578b 100644
--- a/MindIE/MultiModal/StepVideo-TI2V/README.md
+++ b/MindIE/MultiModal/StepVideo-TI2V/README.md
@@ -56,7 +56,7 @@ pip install torch_npu-{pytorchversion}.xxxx.{arch}.whl
 ## 二、下载权重
 
 ### 2.1 权重及配置文件说明
-|     Models      |                        🤗 Huggingface                         |                         MODELERS                          |
+|     Models      |                        Huggingface                         |                         MODELERS                          |
 | :-------------: | :----------------------------------------------------------: | :----------------------------------------------------------: |
 | Step-Video-TI2V | [Download](https://huggingface.co/stepfun-ai/stepvideo-ti2v) | [Download](https://modelers.cn/models/StepFun/Step-Video-TI2V-npu-weight) |
 
-- 
Gitee