Created by: putcn
Purpose
This is an automation tool for deploying paddlepaddle benchmark testing to aws.
Features
- subnet creation to fit just the amount of ec2 instances required.
- pserver and trainer ec2 instances allocation, and instance state verification
- nvidia-docker ready for GPU training
- Instances and network element garbage collection when a task is accomplished or an error occurred
- Test log is collected in realtime
- Web service for checking log or tearing down the testing setup
- No testing code change needed
- Lots of optional configuration options
for more info, please refer to the README.md in this PR.