v1.1

README.md

@@ -47,8 +47,8 @@ data
         ...
     val
         |
-        n04286575
-        n04596742
+        n01440764
+        n01824575
         ...
     test
         |
@@ -87,13 +87,11 @@ export LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/public/home/xxx/anaconda3/envs/megatr
 ```
 **单机多卡**（需先单独申请线上节点）：
 ```
-cd examples
-sh dspvit_1node.sh
+sh examples/dspvit_1node.sh
 ```
 **单机单卡**（需先单独申请线上节点）：
 ```
-cd examples
-dspvit_1dcu.sh
+sh examples/dspvit_1dcu.sh
 ```
 ### 二、mpirun训练
 注释[`arguments.py`](./megatron/arguments.py)中的rank和world_size：

docker/Dockerfile

+FROM image.sourcefind.cn:5000/dcu/admin/base/pytorch:1.10.0-centos7.6-dtk-23.04-py38-latest
+ENV DEBIAN_FRONTEND=noninteractive
+# RUN yum update && yum install -y git cmake wget build-essential
+RUN source /opt/dtk-23.04/env.sh
+# 安装pip相关依赖
+COPY requirements.txt requirements.txt
+RUN pip3 install -i http://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com -r requirements.txt

docker/requirements.txt

+datasets
+nltk
+numpy==1.23.5
+parameterized
+pybind11
+regex
+six
+tensorboard
+transformers
+# versions from HF transformers
+black==21.4b0
+isort>=5.5.4
+ninja

model.properties

 # 模型编码
-modelCode=342
+modelCode=360
 # 模型名称
 modelName=megatron-deepspeed-vit_pytorch
 # 模型描述
 modelDescription=基于transformer的图像分类算法
 # 应用场景
-appScenario=推理,训练,图像分类
+appScenario=推理,训练,图像分类,制造,环境,医疗,气象
 # 框架类型
 frameType=PyTorch

requirements.txt

 datasets
 nltk
-numpy
+numpy==1.23.5
 parameterized
 pybind11
 regex