ATOM P2: ship MI325X lab-validated configs and drop engine-blocked models - #381

Merged

hnimra-amd merged 36 commits into

mainfrom

hnimrama/atom_p2

Sep 9, 2026

Commits on Sep 2, 2026

Add vllm/sglang parity profiles to ATOM single-node configs.
hnimra-amd
committed
a972799
Fix ATOM lab run blockers in pytest and run-deck hooks.
hnimra-amd
committed
37ddc70
Relax OpenAI smoke JSON checks for Qwen and GLM thinking models.
hnimra-amd
committed
433bec1
Fix lifecycle HTML table formatting for non-float metric values.
hnimra-amd
committed
a03a624
Fix native ATOM server args for MoE models and drop GLM 5.1 configs.
hnimra-amd
committed
184aafe

Commits on Sep 8, 2026

Commits on Sep 9, 2026