The PyTorch multi-cloud working group has been created to develop a sustainable, equitable, community managed approach to the CI/CD pipeline for PyTorch in a multi-cloud environment.
This working group complement the work done by the PyTorch CI Working Group by focusing on expanding the CI/CD infrastructure to a multi-cloud environment, in a way that is sustainable for the community and for the CI WG.
The working group initial list of objectives are:
- Develop a multi-cloud architecture with cloud-agnostic monitoring and observability.
- Leverage cloud credits to maintain test coverage without increasing infrastructure spend
- Extend the list of node-providers to improve test coverage without increasing infrastructure spend
Once these objectives will be achieved, the group will evaluate whether to continue its activity beyond these initial objectives.
The group will produce the following artifacts and outcomes:
- Solution design and presentation
- Implementation plan
- Staffing and implementation of the solution
As artifacts are produced, they will be stored in this repositories.
Documents will be in the docs folder.
Working group members, in alphabetical order, are:
- Andrea Frittoli, IBM (lead)
- Andrey Talman, Meta
- Ankit Patel, NVIDIA
- Chen Zeng, Huawei
- Chris Sosa, AMD
- Eli Uriegas, Meta
- Israel Arnar, NVIDIA
- Jean Schmidt, Meta
- Jeremy Eder, RedHat
- Jessica Gonzalez, Linux Foundation
- Jitendra Patil, Intel
- Kiuk Chung, Google
- Balaji Ethirajulu, Linux Foundation
- Scott Suchyta, NVIDIA
- Thanh Ha, Linux Foundation
- Regina Nkenchor, Linux Foundation
- Zain Rizvi, Meta
- Zhe Zhang, NVIDIA
New members are welcome! Membership to this Working Group is open to public and self-declared. To join, please:
- Reach out to Andrea Frittoli on the PyTorch slack
- Make a PR adding your name to the list of members
- Read the available documentation
- Join the meetings on a regular basis!
The Working Groups meets every week on Thursday at 5pm UTC. Please reach out to Andrea Frittoli to get an invite. Meetings are recorded. Notes are recordings are publicly available.
- Slack: Reach out to Andrea Frittoli on the PyTorch slack to get an invite to the WG channel