Allow the Suggested Reviewer scenario to test model training features
E2E Testing for Suggested Reviewer is being implemented in !1024 (merged)
Those tests depend on a pretrained model. Model training requires separate infrastructure and more compute resources (including GPUs).
This issue concerns extending the e2e tests to include training. The aim is to reproduce the end-to-end workflow users would have to implement to be able to deploy Suggested Reviewers in a self-managed environment.
Phase 1 - Research
-
Identify any existing test/benchmark/demo model training that could be used. -
Investigate what work is needed to adapt existing tools for use by GitLab QA, or to create it from scratch.
Phase 2 - Implementation
- Implement model training in GitLab QA
Edited by Mark Lapierre