diff --git a/PyTorch/built-in/mlm/HunyuanDiT/README.md b/PyTorch/built-in/mlm/HunyuanDiT/README.md
index 1c216c658ba5271de1a8925fbf7e1f9604be94e3..1a29bea836c9af2041932668869725bd117efc8e 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/README.md
+++ b/PyTorch/built-in/mlm/HunyuanDiT/README.md
@@ -2,13 +2,13 @@
 # HunyuanDiT for PyTorch
 # 目录
 
-- [HunyuanDiT for PyTorch](#opensora11-for-pytorch)
+- [HunyuanDiT for PyTorch](#HunyuanDiT-for-pytorch)
 - [目录](#目录)
 - [简介](#简介)
   - [模型介绍](#模型介绍)
   - [支持任务列表](#支持任务列表)
   - [代码实现](#代码实现)
-- [HunyuanDiT（在研版本）](#stdit2在研版本)
+- [HunyuanDiT（在研版本）](#HunyuanDiT在研版本)
   - [准备训练环境](#准备训练环境)
     - [安装模型环境](#安装模型环境)
     - [安装昇腾环境](#安装昇腾环境)
@@ -85,7 +85,6 @@ HunyuanDiT是由腾讯开发并开源的一款先进的文生图（文本到图
    ```python
    source ${cann_install_path}/ascend-toolkit/set_env.sh              # 激活cann环境
    cd HunyuanDiT
-   pip install -v -e .                                                # 安装本地代码仓
    pip install -r requirements.txt                                    #安装其它依赖
    ```
 
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
index aa7b4e9bc446ad0e74130c33dee64f11fa906201..12860bdde2bb9352906577f5e984414323eb6bef 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
@@ -32,6 +32,7 @@ from utils.npu_utils import is_npu_available
 if is_npu_available():
     import torch_npu
     from torch_npu.contrib import transfer_to_npu
+    torch.npu.config.allow_internal_format = False
 
 
 class Resolution:
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
index 3b7245bbcf49c2e645d265cc60558d9a986b70cc..551527de9215615d02e9fe4c53366f3e36b0cbba 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
@@ -43,6 +43,7 @@ if is_npu_available():
     import torch_npu
     from torch_npu.npu.amp import autocast
     from torch_npu.contrib import transfer_to_npu
+    torch.npu.config.allow_internal_format = False
 else:
     from torch.cuda.amp import autocast
 
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt b/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
index ba84538f16a8010a2a3b8092691ddf56ffd0c876..2a44c47d35a13a530ed564afa14fa7f2cfaaa045 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
+++ b/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
@@ -11,4 +11,7 @@ sentencepiece==0.1.99
 nvidia-pyindex==1.0.9
 pandas==2.0.3
 gradio==3.50.2
-deepspeed==0.14.4
\ No newline at end of file
+deepspeed==0.14.4
+pyarrow==16.1.0
+decorator
+scipy
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/test/train_full_8p_bf16.sh b/PyTorch/built-in/mlm/HunyuanDiT/test/train_full_8p_bf16.sh
index fe0fb6048a78ff1ce593ea3e37d5f39192aeca28..dc682c2e28f4338f88933be8d75a62252092240b 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/test/train_full_8p_bf16.sh
+++ b/PyTorch/built-in/mlm/HunyuanDiT/test/train_full_8p_bf16.sh
@@ -104,8 +104,10 @@ CaseName=${Network}_bs${BatchSize}_${WORLD_SIZE}'p'_'acc'
 # 结果打印，不需要修改
 echo "------------------ Final result ------------------"
 # 输出性能FPS，需要模型审视修改
-avg_time =`grep -a 'Steps/Sec:'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk -F "Steps/Sec: " '{print $2}' | awk '{a+=$1} END {if (NR != 0) printf("%.3f",a/NR)}'`
-FPS=`echo "$avg_time * $BatchSize" |bc`
+avg_time=$(grep -a 'Steps/Sec:' "${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log" |
+           awk -F "Steps/Sec:" '{print $2}' |
+           awk '{a+=$1} END {if (NR!=0) printf "%.3f\n", a/NR}')
+FPS=$(echo "$avg_time * $BatchSize" |bc)
 
 # 打印，不需要修改
 echo "Final Performance images/sec : $FPS"