MindWizard 文档

mindwizard [-h] [--version] name
positional arguments:
  name        Specify the new project name.
optional arguments:
  -h, --help  show this help message and exit
  --version   show program's version number and exit

project
 |- script
 |   |- run_standalone_train.sh     # 单卡训练脚本
 |   |- run_distribute_train.sh     # 多卡训练脚本
 |   |- run_eval.sh                 # 评估脚本
 |   |- ...
 |- src
 |   |- config.py                   # 参数配置
 |   |- dataset.py                  # 数据集处理
 |   |- lenet.py/resent.py/...      # 网络定义
 |   |- ...
 |- eval.py                         # 网络评估
 |- train.py                        # 网络训练
 |- README.md

$ mindwizard project

>>> Please select a network:
   1: alexnet
   2: lenet
   3: resnet50
 : 2
Your choice is lenet.
>>> Please select a loss function:
   1: SoftmaxCrossEntropyWithLogits
 [1]: 1
Your choice is SoftmaxCrossEntropyWithLogits.
>>> Please select an optimizer:
   1: Adam
   2: Momentum
   3: SGD
 [2]: 2
Your choice is Momentum.
>>> Please select a dataset:
   1: MNIST
 [1]: 1
Your choice is MNIST.

project is generated in $PWD/project

$ cd $PWD/project/scripts

# Ascend
Usage: bash run_distribute_train.sh [RANK_TABLE_FILE] [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)

# GPU
Usage: bash run_distribute_train_gpu.sh [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)

# Ascend
Usage: bash run_standalone_train.sh [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)

# GPU 
Usage: bash run_standalone_train_gpu.sh [DATASET_PATH] [PRETRAINED_CKPT_PATH](optional)

# Ascend
Usage: bash run_eval.sh [DATASET_PATH] [CHECKPOINT_PATH]

# GPU
Usage: bash run_eval_gpu.sh [DATASET_PATH] [CHECKPOINT_PATH]