ModuleNotFoundError:运行nvidia/deeplearningexamples repostory时，没有名为“nvidia.dali.backend_impl”的模块

Question

在NVIDIA/DeepLearningExamples中运行rn50脚本，如下所示

./rn50_partial.sh

获取如下错误：

Traceback (most recent call last):

  File "/export/nfs/sunxue/DeepLearningExamples/PyTorch/Classification/ConvNets/mnasnet/training/FP32/../../../launch.py", line 7, in <module>

    from main import main, add_parser_arguments, available_models

  File "/export/nfs/sunxue/DeepLearningExamples/PyTorch/Classification/ConvNets/main.py", line 49, in <module>

    from image_classification.dataloaders import *

  File "/export/nfs/sunxue/DeepLearningExamples/PyTorch/Classification/ConvNets/image_classification/dataloaders.py", line 79, in <module>

    class HybridTrainPipe(Pipeline):

NameError: name 'Pipeline' is not defined

Answer 1

好的，我已经解决了这个问题，因为NVIDIA没有对具体的安装进行分类，实际上，它不必从头开始安装，如果您使用pip安装，只需检查NVIDIA docker使用的版本，然后下载这个特定的版本，pip安装，一切都会好的