diff --git a/PyTorch/built-in/mlm/HunyuanDiT/README.md b/PyTorch/built-in/mlm/HunyuanDiT/README.md
index b40bae9818a5e205aa43c2241adb577ff5b9345a..e06d76af2f5c684b64ecb89692154fc0cf150ccd 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/README.md
+++ b/PyTorch/built-in/mlm/HunyuanDiT/README.md
@@ -28,8 +28,8 @@
 # 简介
 ## 模型介绍
 
-OpenSora是HPC AI Tech开发的开源高效复现类Sora视频生成方案。OpenSora不仅实现了先进视频生成技术的低成本普及，还提供了一个精简且用户友好的方案，简化了视频制作的复杂性。
-本仓库主要将OpenSora1.1的STDiT2模型的任务迁移到了昇腾NPU上，并进行极致性能优化。
+HunyuanDiT是由腾讯开发并开源的一款先进的文生图（文本到图像）模型。该模型支持中英文双语输入，特别针对中文进行了优化，能够深刻理解中文语境和文化元素，生成高质量且富有中国文化特色的图像。HunyuanDiT经过大规模中文数据集的训练，涵盖了广泛的类别和艺术风格，能够根据文本提示生成细腻逼真的图像。
+本仓库主要将HunyuanDiT模型的任务迁移到了昇腾NPU上，并进行极致性能优化。
 
 ## 支持任务列表
 
@@ -48,14 +48,14 @@ OpenSora是HPC AI Tech开发的开源高效复现类Sora视频生成方案。Ope
 
   ```
   url=https://github.com/Tencent/HunyuanDiT
-  commit_id=(待补充)
+  commit_id=3bb80e1dedba5bf9728e7c9566c4b5c665bbfbd2
   ```
 
 - 适配昇腾 AI 处理器的实现：
 
   ```
   url=https://gitee.com/ascend/ModelZoo-PyTorch.git
-  code_path=PyTorch/built-in/mlm/
+  code_path=PyTorch/built-in/mlm/HunyuanDiT
   ```
 
 
@@ -71,7 +71,12 @@ OpenSora是HPC AI Tech开发的开源高效复现类Sora视频生成方案。Ope
   |     三方库     |  支持版本  |
   |:-----------:|:------:|
   |   PyTorch   | 2.1.0  |
-  | TorchVision | 0.16.0 |
+  | TorchVision | 0.14.1 |
+  | deepspeed | 0.14.4 |
+  | diffusers | 0.21.2 |
+| transformers | 4.39.1 |
+| accelerate | 0.27.2 |
+
 
 
    在模型根目录下执行以下命令，安装模型对应PyTorch版本需要的依赖。
@@ -79,8 +84,9 @@ OpenSora是HPC AI Tech开发的开源高效复现类Sora视频生成方案。Ope
 
    ```python
    source ${cann_install_path}/ascend-toolkit/set_env.sh              # 激活cann环境
-   cd OpenSora1.1
-   pip install -v -e .                                                # 安装本地代码仓，同时自动安装依赖
+   cd HunyuanDiT
+   pip install -v -e .                                                # 安装本地代码仓
+   pip install -r requirements.txt                                    #安装其它依赖
    ```
 
 ### 安装昇腾环境
@@ -97,7 +103,6 @@ OpenSora是HPC AI Tech开发的开源高效复现类Sora视频生成方案。Ope
   | 昇腾NPU固件 |   在研版本   | 
   | 昇腾NPU驱动 | 在研版本 |
 
-  
 
 ### 准备数据集
 #### 训练数据集准备
@@ -110,23 +115,16 @@ https://github.com/Tencent/HunyuanDiT
 
 2. 无网络时，用户可访问huggingface官网自行下载(https://huggingface.co/Tencent-Hunyuan/HunyuanDiT/tree/main)
 
-
-3. 获取对应的预训练模型后，在以下配置文件中将`model`、`vae`的`from_pretrained`参数设置为本地预训练模型绝对路径。
-   ```shell
-   configs/opensora-v1-1/inference/sample.py
-   configs/opensora-v1-1/train/stage1.py
-   configs/opensora-v1-1/train/stage2.py
-   configs/opensora-v1-1/train/stage3.py
+3. 将下载好的t5模型放在本工程目录下的`ckpts`目录下，组织结构如下：
    ```
-
-4. 将下载好的t5模型放在本工程目录下的`DeepFloyd`目录下，组织结构如下：
-   ```
-   $OpenSora1.1
-   ├── DeepFloyd
-   ├── ├── t5-v1_1-xxl
-   ├── ├── ├── config.json
-   ├── ├── ├── pytorch_model-00001-of-00002.bin
-   ├── ├── ├── ...
+   $HunyuanDiT
+   ├── ckpts
+   ├── ├── t2i
+   ├── ├── ├── clip_text_encoder
+   ├── ├── ├── model
+   ├── ├── ├── mt5
+   ├── ├── ├── sdxl-vae-fp16-fix
+   ├── ├── ├── tokenizer
    └── ...
    ```
 
@@ -159,24 +157,12 @@ https://github.com/Tencent/HunyuanDiT
     │  │  ├──porcelain_mt.json
    ```
 
-2. 运行训练脚本（待补充）。
-
-   用户可以按照自己训练需要进行参数配置，以下给出单卡和多卡的一种训练示例。
-   ```shell
-   bash test/train_full_1p_opensorav1_1.sh --data_path=train_data.csv
-   # 混合精度BF16，单卡训练，stage1
-   ```
+2. 运行训练脚本。
 
+   用户可以按照自己训练需要进行参数配置，以下给出多卡的一种训练示例。
    ```shell
-   bash test/train_full_8p_opensorav1_1.sh --data_path=train_data.csv
-   # 混合精度BF16，八卡训练，stage1
-   ```
-   对于本模型，可以采用绑核优化，以绑核方式启动。
-   绑核方法参考：https://gitee.com/ascend/att/tree/master/profiler/affinity_cpu_bind
-   本模型使用示例如下：
-   ```
-   python3 bind_core.py \
-   -app="bash test/train_full_18p_opensorav1_1.sh --data_path=train_data.csv"
+   bash test/train_full_8p_bf16.sh
+   # 混合精度BF16，8卡训练
    ```
 
 ### 推理任务
@@ -193,35 +179,14 @@ https://github.com/Tencent/HunyuanDiT
 
 - 单机单卡推理
   ```shell
-  bash test/infer_full_1p_opensorav1_1.sh --ckpt_path=/path/to/OpenSora-STDiT-v2-stage3/model.pth  # 混精bf16 在线推理
+  bash test/inference_full_1p_fp16.sh  # 混精fp16 在线推理
   ```
 - 推理脚本参数说明如下
    ```shell
-   test/infer_full_1p_opensorav1_1.sh
-   --batch_size                         //设置batch_size
-   --ckpt_path                          //推理加载的模型地址
-   --prompt                             //测试用的prompt
-   --num_frames                         //生成视频的总帧数
-   --img_h                              //生成视频的宽
-   --img_w                              //生成视频的高
-  
-   scripts/inference.py
-   config                               //配置文件路径
-   --seed                               //随机种子
-   --ckpt-path                          //推理加载的模型文件路径    
-   --batch-size                         //设置batch_size
-   --prompt-path                        //推理使用的prompt文件路径
-   --prompt                             //测试用的prompt
-   --num-frames                         //生成视频的总帧数
-   --image-size                         //生成视频的分辨率
-   --fps                                //生成视频的帧率
-   --save-dir                           //输出视频的路径
-   --num-sampling-steps                 //推理的采样步数
-   --cfg-scale                          //无分类器引导的权重系数
+   test/inference_full_1p_fp16.sh
+   --prompt                         //测试用的prompt
    ```
 
-
-
 # 公网地址说明
 代码涉及公网地址参考 public_address_statement.md
 
@@ -229,7 +194,7 @@ https://github.com/Tencent/HunyuanDiT
 
 ## 变更
 
-2024.04.29：OpenSora1.1 STDiT2 bf16训练和推理任务首次发布。
+2024.08.22：HunyuanDiT bf16训练和fp16推理任务首次发布。
 
 # FAQ
 
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/gpu_sample.sh b/PyTorch/built-in/mlm/HunyuanDiT/gpu_sample.sh
deleted file mode 100644
index f7721e48aa57c29a43307c8d61075bb89412e2d3..0000000000000000000000000000000000000000
--- a/PyTorch/built-in/mlm/HunyuanDiT/gpu_sample.sh
+++ /dev/null
@@ -1 +0,0 @@
-python sample_t2i.py --infer-mode fa --prompt "渔舟唱晚" --image-size 1280 768 --no-enhance
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/gpu_train.sh b/PyTorch/built-in/mlm/HunyuanDiT/gpu_train.sh
deleted file mode 100644
index 95005fbcdcc032dfd647d57b054500c435722772..0000000000000000000000000000000000000000
--- a/PyTorch/built-in/mlm/HunyuanDiT/gpu_train.sh
+++ /dev/null
@@ -1,2 +0,0 @@
-export CUBLAS_WORKSPACE_CONFIG=:4096:8
-PYTHONPATH=./ sh hydit/train.sh
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
index 682f29d7faa2d17a567e9f4d623f6eb8e1361181..6470cb6356362a9af95739d5c25293bf176cdb52 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/inference.py
@@ -23,6 +23,12 @@ from .modules.text_encoder import MT5Embedder
 from .utils.tools import set_seeds
 from peft import LoraConfig
 
+from utils.npu_utils import is_npu_available
+
+if is_npu_available():
+    import torch_npu
+    from torch_npu.contrib import transfer_to_npu
+
 
 class Resolution:
     def __init__(self, width, height):
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/attn_layers.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/attn_layers.py
index ba666e74d93e0d905f03b74b56aacc452877a272..7ddfbafd071f734b53cec3f534a9775526f91888 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/attn_layers.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/attn_layers.py
@@ -2,6 +2,7 @@ import torch
 import torch.nn as nn
 from typing import Tuple, Union, Optional
 
+import math
 from utils.npu_utils import is_npu_available
 if is_npu_available():
     import torch_npu
@@ -148,7 +149,7 @@ class FlashSelfMHAModified(nn.Module):
                  device=None,
                  dtype=None,
                  norm_layer=nn.LayerNorm,
-                 FAG_deterministic=False,
+                 fa_deterministic=False,
                  ):
         factory_kwargs = {'device': device, 'dtype': dtype}
         super().__init__()
@@ -167,7 +168,7 @@ class FlashSelfMHAModified(nn.Module):
         if is_npu_available():
             self.inner_attn = NpuFlashAttention(attention_dropout=attn_drop)
         else:
-            self.inner_attn = FlashSelfAttention(attention_dropout=attn_drop, deterministic=FAG_deterministic)
+            self.inner_attn = FlashSelfAttention(attention_dropout=attn_drop, deterministic=fa_deterministic)
         self.out_proj = nn.Linear(dim, dim, bias=qkv_bias, **factory_kwargs)
         self.proj_drop = nn.Dropout(proj_drop)
 
@@ -230,7 +231,7 @@ class FlashCrossMHAModified(nn.Module):
                  device=None,
                  dtype=None,
                  norm_layer=nn.LayerNorm,
-                 FAG_deterministic=False,
+                 fa_deterministic=False,
                  ):
         factory_kwargs = {'device': device, 'dtype': dtype}
         super().__init__()
@@ -255,7 +256,7 @@ class FlashCrossMHAModified(nn.Module):
         if is_npu_available():
             self.inner_attn = NpuFlashAttention(attention_dropout=attn_drop)
         else:
-            self.inner_attn = FlashCrossAttention(attention_dropout=attn_drop, deterministic=FAG_deterministic)
+            self.inner_attn = FlashCrossAttention(attention_dropout=attn_drop, deterministic=fa_deterministic)
 
         self.out_proj = nn.Linear(qdim, qdim, bias=qkv_bias, **factory_kwargs)
         self.proj_drop = nn.Dropout(proj_drop)
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/models.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/models.py
index 69dfc4384f1ecf579e8ac32a91c264d21ac04573..b0b20472d53848de33925863a1767bee4e6a2013 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/models.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/modules/models.py
@@ -326,7 +326,7 @@ class HunYuanDiT(ModelMixin, ConfigMixin, PeftAdapterMixin):
 
         # Build image meta size tokens
         image_meta_size = timestep_embedding(image_meta_size.view(-1), 256)   # [B * 6, 256]
-        if self.args.use_fp16 and self.args.autocast_dtype == "fp16":
+        if self.args.use_fp16:
             image_meta_size = image_meta_size.half()
         image_meta_size = image_meta_size.view(-1, 6 * 256)
         extra_vec = torch.cat([extra_vec, image_meta_size], dim=1)  # [B, D + 6 * 256]
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/run_g.sh b/PyTorch/built-in/mlm/HunyuanDiT/hydit/run_g.sh
index 63a9e30f9fe3f30f334d2f976f528be3d94628af..f145b56937b9649a888090b9ce53026831563d06 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/run_g.sh
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/run_g.sh
@@ -6,26 +6,3 @@ params=" \
             --rope-real \
             "
 deepspeed --num_gpus 8 --num_nodes 1 --master_port=29000 hydit/train_deepspeed.py ${params}  "$@"
-
-#HOSTFILE="/home/l50041210/HunyuanDiT_combine/hostfile"
-#MASTER_ADDR=$(head -n1 $HOSTFILE | awk '{print $1;}')
-#MASTER_PORT=6001
-#NODE_ADDR=`hostname -I | awk '{for(i=1;i<=NF;i++)print $i}' | grep ${MASTER_ADDR%.*}. | awk -F " "'{print$1}'`
-#NODE_RANK=$(awk '{ranks[$1]=(FNR-1);}END{print ranks["'$NODE_ADDR'"];}' $HOSTFILE)
-#NNODES=$(cat $HOSTFILE | wc -l)
-#NPUS_PER_NODE=8
-#WORLD_SIZE=$(($NPUS_PER_NODE*$NNODES))
-#echo $MASTER_ADDR
-#echo $NODE_ADDR
-#echo $NODE_RANK
-#echo $NNODES
-#
-#DISTRIBUTED_ARGS="
-#    --nproc_per_node $NPUS_PER_NODE \
-#    --nnodes $NNODES \
-#    --node_rank $NODE_RANK \
-#    --master_addr $MASTER_ADDR \
-#    --master_port $MASTER_PORT
-##"
-#
-#torchrun $DISTRIBUTED_ARGS hydit/train_deepspeed.py ${params} "$@"
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train.sh b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train.sh
index 10c856dfe1064716d2227a4ca53385da709c5160..c6c235d2688394874fe2a35ff1678890f4aab372 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train.sh
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train.sh
@@ -43,4 +43,4 @@ sh $(dirname "$0")/run_g.sh \
     --max-training-steps 5000 \
     --norm 'layer' \
     --autocast-dtype 'bf16'
-    "$@"
\ No newline at end of file
+    "$@"
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
index 05ce0655f821dbb21d80d3d04297f43b61020f82..c82c29232cd4d9508cd65824c2d29bb1bb3661f1 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
+++ b/PyTorch/built-in/mlm/HunyuanDiT/hydit/train_deepspeed.py
@@ -539,7 +539,5 @@ def main(args):
 
 if __name__ == "__main__":
     args = get_args()
-    if args.seed_all:
-        seed_all(is_gpu=not is_npu_available(), mode=True)
     # Start
     main(args)
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/npu_sample.sh b/PyTorch/built-in/mlm/HunyuanDiT/npu_sample.sh
deleted file mode 100644
index df4d8f0fa7c2c440e622a217f13e4a57517d5a14..0000000000000000000000000000000000000000
--- a/PyTorch/built-in/mlm/HunyuanDiT/npu_sample.sh
+++ /dev/null
@@ -1,8 +0,0 @@
-source /home/l50041210/cann-b20/ascend-toolkit/set_env.sh
-export ASCEND_RT_VISIBLE_DEVICES=0
-#将Host日志输出到串口,0-关闭/1-开启
-export ASCEND_SLOG_PRINT_TO_STDOUT=0
-#设置默认日志级别,0-debug/1-info/2-warning/3-error
-export ASCEND_GLOBAL_LOG_LEVEL=3
-
-python sample_t2i.py --infer-mode fa --prompt "渔舟唱晚" --image-size 1280 768 --no-enhance
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/npu_train.sh b/PyTorch/built-in/mlm/HunyuanDiT/npu_train.sh
deleted file mode 100644
index 57e991e0dc377a14ab7af0f5b3db3763b123b3cb..0000000000000000000000000000000000000000
--- a/PyTorch/built-in/mlm/HunyuanDiT/npu_train.sh
+++ /dev/null
@@ -1,7 +0,0 @@
-source /home/l50041210/cann-b020/ascend-toolkit/set_env.sh
-#将Host日志输出到串口,0-关闭/1-开启
-export ASCEND_SLOG_PRINT_TO_STDOUT=0
-#设置默认日志级别,0-debug/1-info/2-warning/3-error
-export ASCEND_GLOBAL_LOG_LEVEL=3
-
-PYTHONPATH=./ sh hydit/train.sh
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/public_address_statement.md b/PyTorch/built-in/mlm/HunyuanDiT/public_address_statement.md
new file mode 100644
index 0000000000000000000000000000000000000000..226cffb32fe8b808c20bcc91d1a82afb71a5b3c6
--- /dev/null
+++ b/PyTorch/built-in/mlm/HunyuanDiT/public_address_statement.md
@@ -0,0 +1,26 @@
+| 类型 | 开源代码地址                                                     | 文件名                             | 公网IP地址/公网URL地址/域名/邮箱地址 | 用途说明                |
+| ------- |------------------------------------------------------------|---------------------------------|------------------------|---------------------|					
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/utils/collect_env.py	 | 	./utils/collect_env.py		       |  	    https://github.com/open-mmlab/mmengine/issues/931               | 	注释参考链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/mllm/llava/utils.py  | 	./mllm/llava/utils.py		        | 	      https://api.openai.com/v1/moderations                | 	文件交互链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/mllm/llava/model/builder.py	                 | 	./mllm/llava/model/builder.py	 |  	        https://github.com/haotian-liu/LLaVA#launch-a-model-worker-lora-weights-unmerged             | 	模型权重公网下载说明	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/lite/inference.py	                 | 		         ./lite/inference.py  |  	  https://huggingface.co/Tencent-Hunyuan                    | 	huggingface官仓说明链接	 |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/modules/posemb_layers.py	                 | 			           hydit/modules/posemb_layers.py                  |  	    https://github.com/facebookresearch/llama/blob/main/llama/model.py                  | 	注释参考链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/modules/posemb_layers.py	                 | 			               hydit/modules/posemb_layers.py              |  	        https://github.com/facebookresearch/mae/blob/main/util/pos_embed.py              | 	注释参考链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/modules/embedders.py	                 | 			         hydit/modules/embedders.py                    |  	       https://github.com/openai/glide-text2im/blob/main/glide_text2im/nn.py               | 	注释参考链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/modules/embedders.py	                 | 			         hydit/modules/embedders.py                    |  	       https://github.com/google-research/vision_transformer               | 	注释参考链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/lr_scheduler.py	                 | 			     hydit/lr_scheduler.py                        |  	      https://arxiv.org/abs/1803.09820                 | 	参考论文链接	            |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			    hydit/diffusion/pipeline_controlnet.py                         |  	       https://arxiv.org/pdf/2305.08891.pdf               | 	参考论文链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			    hydit/diffusion/pipeline_controlnet.py                         |  	      https://arxiv.org/abs/2010.02502                | 	参考论文链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			   hydit/diffusion/pipeline_controlnet.py                          |  	     https://arxiv.org/pdf/2205.11487.pdf                 | 	参考论文链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			  hydit/diffusion/pipeline_controlnet.py                           |  	    https://huggingface.co/openai/clip-vit-large-patch14                  | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			  hydit/diffusion/pipeline_controlnet.py                           |  	   https://huggingface.co/runwayml/stable-diffusion-v1-5                   | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			  hydit/diffusion/pipeline_controlnet.py                           |  	  https://github.com/huggingface/diffusers/pull/254                    | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			 hydit/diffusion/pipeline_controlnet.py                            |  	 https://pytorch.org/docs/stable/generated/torch.Generator.html                     | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/hydit/diffusion/pipeline_controlnet.py	                 | 			 hydit/diffusion/pipeline_controlnet.py                            |  	https://github.com/huggingface/diffusers/blob/main/src/diffusers/models/attention_processor.py                      | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/diffusion/gaussian_diffusion.py                     | 			    comfyui-hydit/hydit/diffusion/gaussian_diffusion.py                         |  	      https://github.com/hojonathanho/diffusion/blob/1e0dceb3b3495bbe19116a5e1b3596cd0706c543/diffusion_tf/diffusion_utils_2.py                | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/diffusion/gaussian_diffusion.py	                    | 			comfyui-hydit/hydit/diffusion/gaussian_diffusion.py                             |  	 https://www.crosslabs.org/blog/diffusion-with-offset-noise                     | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/diffusion/gaussian_diffusion.py                     | 			    comfyui-hydit/hydit/diffusion/gaussian_diffusion.py                         |  	    https://openreview.net/forum?id=PlKWVd2yBkY                  | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/annotator/glyph.py	                    | 			 comfyui-hydit/hydit/annotator/glyph.py                           |  	      https://github.com/AIGText/GlyphControl-release                | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/annotator/dwpose/util.py	                    | 			  comfyui-hydit/hydit/annotator/dwpose/util.py                          |  	       https://github.com/CMU-Perceptual-Computing-Lab/openpose/blob/master/src/openpose/hand/handDetector.cpp               | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/annotator/dwpose/__init__.py                     | 			     comfyui-hydit/hydit/annotator/dwpose/__init__.py                        |  	   https://github.com/CMU-Perceptual-Computing-Lab/openpose                   | 	注释参考链接	        |
+|	开源代码引入	| https://github.com/Tencent/HunyuanDiT/comfyui-hydit/hydit/annotator/dwpose/__init__.py	                    | 			  comfyui-hydit/hydit/annotator/dwpose/__init__.py                           |  	https://github.com/Hzzone/pytorch-openpose                      | 	注释参考链接	        |
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt b/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
index 8de35fd32eb220912bf782b94dad64beb2776e20..ba84538f16a8010a2a3b8092691ddf56ffd0c876 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
+++ b/PyTorch/built-in/mlm/HunyuanDiT/requirements.txt
@@ -1,22 +1,14 @@
---extra-index-url https://pypi.ngc.nvidia.com
 timm==0.9.5
 diffusers==0.21.2
-peft==0.10.0
+peft==0.11.1
 protobuf==3.19.0
 torchvision==0.14.1
 transformers==4.39.1
-peft==0.10.0
-accelerate==0.29.3
+accelerate==0.27.2
 loguru==0.7.2
 einops==0.7.0
 sentencepiece==0.1.99
-cuda-python==11.7.1
-onnxruntime==1.12.1
-onnx==1.12.0
 nvidia-pyindex==1.0.9
-onnx-graphsurgeon==0.3.27
-polygraphy==0.47.1
 pandas==2.0.3
 gradio==3.50.2
-deepspeed==0.6.3
-pyarrow==16.1.0
\ No newline at end of file
+deepspeed==0.14.4
\ No newline at end of file
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/test/env_npu.sh b/PyTorch/built-in/mlm/HunyuanDiT/test/env_npu.sh
index cd24c60a82fa5e910db8306e559579206d7465d9..a2c0d400c514eb870c15222c73906ddea8dae0b5 100644
--- a/PyTorch/built-in/mlm/HunyuanDiT/test/env_npu.sh
+++ b/PyTorch/built-in/mlm/HunyuanDiT/test/env_npu.sh
@@ -33,4 +33,4 @@ export COMBINED_ENABLE=1
 #HCCL白名单开关,1-关闭/0-开启
 export HCCL_WHITELIST_DISABLE=1
 export HCCL_IF_IP=$(hostname -I |awk '{print $1}')
-export HCCL_CONNECT_TIMEOUT=1200
\ No newline at end of file
+export HCCL_CONNECT_TIMEOUT=1200
diff --git a/PyTorch/built-in/mlm/HunyuanDiT/test/inference_full_1p_fp16.sh b/PyTorch/built-in/mlm/HunyuanDiT/test/inference_full_1p_fp16.sh
new file mode 100644
index 0000000000000000000000000000000000000000..063dbba0e6c91fda2489e2d899586fb6bd082584
--- /dev/null
+++ b/PyTorch/built-in/mlm/HunyuanDiT/test/inference_full_1p_fp16.sh
@@ -0,0 +1,58 @@
+# 网络名称，权重路径以及相关参数，需要模型审视修改
+Network="HunyuanDiT"
+prompt="渔舟唱晚"
+image_size_height=1280
+image_size_weight=768
+
+for para in $*
+do
+    if [[ $para == --prompt* ]]; then
+        prompt=$(echo ${para#*=})
+    fi
+done
+
+# cd到与test文件夹同层级目录下执行脚本，提高兼容性；test_path_dir为包含test文件夹的路径
+cur_path=$(pwd)
+cur_path_last_dirname=${cur_path##*/}
+if [ x"${cur_path_last_dirname}" == x"test" ]; then
+  test_path_dir=${cur_path}
+  cd ..
+  cur_path=$(pwd)
+else
+  test_path_dir=${cur_path}/test
+fi
+
+source ${test_path_dir}/env_npu.sh
+
+#创建DeviceID输出目录，不需要修改
+output_path=${cur_path}/test/output/${ASCEND_DEVICE_ID}
+
+mkdir -p ${output_path}
+
+#推理开始时间，不需要修改
+start_time=$(date +%s)
+echo "start_time: ${start_time}"
+
+python sample_t2i.py --infer-mode fa --prompt ${prompt} --image-size ${image_size_height} ${image_size_weight} --no-enhance \
+         > ${test_path_dir}/output/$ASCEND_DEVICE_ID/inference_${ASCEND_DEVICE_ID}.log 2>&1 &
+
+wait
+
+#推理结束时间，不需要修改
+end_time=$(date +%s)
+e2e_time=$(($end_time - $start_time))
+
+
+# 训练用例信息，不需要修改
+BatchSize=${BATCH_SIZE}
+DeviceType=$(uname -m)
+CaseName=${Network}_bs${BatchSize}_${WORLD_SIZE}'p'_'acc'
+
+# 结果打印，不需要修改
+echo "------------------ Final result ------------------"
+# 打印，不需要修改
+echo "E2E Training Duration sec : $e2e_time"
+
+# 关键信息打印到${CaseName}.log中，不需要修改
+echo "Network = ${Network}" >${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2e_time}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}_perf_report.log
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 596ac19052029386566f1a0b10b81191da787a13..fe0fb6048a78ff1ce593ea3e37d5f39192aeca28 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
@@ -1,4 +1,4 @@
-# 微调生成的ckpt路径
+# 网络名称，权重路径以及相关参数，需要模型审视修改
 Network="HunyuanDiT"
 BATCH_SIZE=1
 max_train_steps=5000
@@ -85,6 +85,8 @@ deepspeed --num_gpus ${WORLD_SIZE} --num_nodes 1 --master_port=${MASTER_PORT} hy
     --multireso \
     --reso-step 64 \
     --max-training-steps ${max_train_steps} \
+    --norm "layer"  \
+    --autocast-dtype  "bf16" \
  > ${test_path_dir}/output/$ASCEND_DEVICE_ID/train_${ASCEND_DEVICE_ID}.log 2>&1 &
 
 wait
@@ -102,7 +104,9 @@ 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
-FPS=`grep -a 'FPS' ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk -F "FPS " '{print $2}' | tail -100 | awk '{a+=$1} END {if (NR != 0) printf("%.3f",a/NR)}'`
+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`
+
 # 打印，不需要修改
 echo "Final Performance images/sec : $FPS"
 echo "E2E Training Duration sec : $e2e_time"
@@ -122,5 +126,6 @@ echo "RankSize = ${WORLD_SIZE}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${Ca
 echo "BatchSize = ${BatchSize}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "DeviceType = ${DeviceType}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "CaseName = ${CaseName}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}_perf_report.log
 echo "TrainingTime = ${TrainingTime}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}_perf_report.log
-echo "E2ETrainingTime = ${e2e_time}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}_perf_report.log
\ No newline at end of file
+echo "E2ETrainingTime = ${e2e_time}" >>${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}_perf_report.log