v0.8.0

Release 0.8.0

Added

• Support using APIs to manage training or serving jobs for Python(arena-python-sdk)
• Support using APIs to manage training or serving jobs for Java(arena-java-sdk)
• Support submitting a seldon serving job
• Support generating the kubeconfig file for the specified user
• Support specifying the starting sequence of the tfjob

Changed

• Refactor the documentation and move documentation to readthedocs
• Reduce execution time of arena
• Remove the deprecated code

Fixed

• Fix the bug of submitting Spark Job
• Fix the bug of viewing logs when chief pod is missing

Please follow the Get started Guide to install.

v0.7.1

• Make et-opertor in arena-system

Please follow the Get started Guide to install.

v0.7.0

• Support using apis to manage training or serving jobs(arena-go-sdk)
• Support getting gpu metrics from Alibaba Cloud ARMS Prometheus
• Support getting node gpu metrics
• Command of "arena get" supports "-g" option
• Support the arena daemon mode, reduce the api-server pressure, arena can listen k8s objects in this mode
• Command of "arena logs" supports "-c" to specify container
• Support to attach a job container and execute some commands("arena attach")
• Command of "arena top node" supports "-r" option

Please follow the Get started Guide to install.

v0.6.0

1. Add Support of Elastic Training, such Elastic Horovod
2. Support using private image

Please follow the Get started Guide to install.

v0.5.0

1. Add Support of Pytorch
2. Add tarball installation for Linux and Mac
3. Support GangScheduling Native in MPIJob

Please follow the Get started Guide to install.

v0.4.0

1. Add GPU support for PS
2. Support Kubernetes 1.18 and above
3. Fix the bug of deploying Prometheus

Please follow the Get started Guide to install.

v0.3.3

1. Support non-root installation
2. Add train init framework
3. Fix the bug of using Estimator

Please follow the Get started Guide to install.

v0.3.2

1. Fix evaluator & chief validation
2. Fix incorrect cpu resource variable, should be psCPU
3. Set exit code as 2 when delete job failed

Please follow the Get started Guide to install.

v0.3.1

1. Upgrade Deployment version from extensions/v1beta1 to apps/v1
2. Fix the issue of incorrect number of allocated GPUs
3. Upgrade Helm to v2.14.1

Please follow the Get started Guide to install.

v0.3.1-beta

Some bugs are fixed.

Please follow the Get started Guide to install.