caffe2

Apache Caffe2 on OpenPAI

This guide introduces how to run Caffe2 job on OpenPAI. The following contents show some basic Caffe2 examples, other customized Caffe2 code can be run similarly.

Caffe2 resnet50 ImageNet example

To run Caffe2 examples in OpenPAI, you need to prepare a job configuration file and submit it through webportal.

OpenPAI packaged the docker env required by the job for user to use. User could refer to DOCKER.md to customize this example docker env. If user have built a customized image and pushed it to Docker Hub, replace our pre-built image openpai/pai.example.caffe2 with your own.

Here's one configuration file example:

resnet50

{
  "jobName": "caffe2-resnet50",
  "image": "openpai/pai.example.caffe2",
  "taskRoles": [
    {
      "name": "main",
      "taskNumber": 1,
      "cpuNumber": 4,
      "memoryMB": 8192,
      "gpuNumber": 1,
      "command": "python resnet50_trainer.py --train_data null"
    }
  ]
}

For more details on how to write a job configuration file, please refer to job tutorial.