Skip to content

capture framework changes in env week over week #3

@robert-s-lee

Description

@robert-s-lee

Background:
Created a ready made reference to address: How do I find out what packages are pre-installed in the cloud machine?
Cloud machines are configured in a simple way with only what is needed to run the scripts in the framework of choice.

Issue:
Not efficient for each new users to run a script when this could be done for all users as a reference material.

Desired outcome:
For each framework lightning|torch|tensorflow|Julia|torchelastic.

  • frameworks/lightning.md
  • frameworks/torch.md
  • frameworks/tensorflow.md
  • frameworks/torchelastic.md

Perhaps follow the categories from GitHub Actions Virtual Environment as a template. The following are minimum information to capture.

  • list current linux packages and versions
  • list current python packages and versions
  • list mounted file systems and sizes
  • list network interfaces and subnets
  • automate the running the check on week over week
  • publish the delta week over week

Solution Approach:
Use Github Actions https://docs.github.com/en/actions/learn-github-actions/understanding-github-actions

When:
once a week (run Sunday)

Where:
Standard_DS2_v2 virtual machines in Microsoft Azure is used to run the test. About GitHub-hosted runners has more info.

Size vCPU Memory: GiB Temp storage (SSD) GiB Max data disks Max cached and temp storage throughput: IOPS/MBps (cache size in GiB) Max uncached disk throughput: IOPS/MBps Max NICs Expected network bandwidth (Mbps)
Standard_DS2_v2 2 7 14 8 8000/64 (86) 6400/96 2 1500

Security:
Grid.ai API is in github secrets. (Register all secrets used within workflows).[https://docs.github.com/en/actions/security-guides/security-hardening-for-github-actions]

Defined the following repository secret

  • GRIDAI_USERNAME
  • GRIDAI_KEY
  • GRIDAI_URL

Metadata

Metadata

Assignees

Labels

No labels
No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions