Skip to content

CI for DDP training #1019

@trivoldus28

Description

@trivoldus28

We need CI for DDP training so that I don't get a heart attack every time I update my branch to main breaking changes to the training code are identified timely and efficiently, both for our code as well as for external dependencies.

To start with, I think just a 15-30m run using 2 nodes every night when there are changes would catch much of the bugs within past few months.

Metadata

Metadata

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions