jgrusewski df4ef104b7 ci: move runner to H100 pool, add training validation stage ...

- Runner node selector: ci-build (L4) → gpu-training (H100) for faster
  compilation (2x+ vCPUs) and native GPU training support
- Resource limits bumped: 20 vCPU / 128Gi (was 10 vCPU / 38Gi)
- New 'train' stage with manual DQN and TFT validation jobs
  (50-step smoke tests to verify CUDA + model correctness)
- Requires helm upgrade of gitlab-runner with new values

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-02-26 18:56:22 +01:00

..

docker

feat: on-demand training dispatch via K8s Jobs with sidecar uploader

2026-02-26 12:43:17 +01:00

k8s

ci: move runner to H100 pool, add training validation stage

2026-02-26 18:56:22 +01:00

live/production

infra(ci): switch CI builds from H100 to L4 — 4.7x cost reduction

2026-02-25 23:30:04 +01:00

modules

infra(ci): switch CI builds from H100 to L4 — 4.7x cost reduction

2026-02-25 23:30:04 +01:00

scripts

refactor(ml): delete 121 dead examples, 6 QAT tests, orphan binary

2026-02-26 00:05:38 +01:00