[AndreasKaratzas]

• kernels.yaml: MI300 MoE Kernels Multi-GPU, and the AITER test_modular_kernel_combinations run in the MI300 MoE Kernels mirror ([ROCm][MoE] Split AiterExperts by weight-quantization key #60322 added both to test-amd.yaml only).
• models_basic.yaml: MI355 Kimi K3 Multi-GPU; its tests skip on the 1-GPU MI355 Kimi K3 mirror.
• expert_parallelism.yaml: a gating MI355 mirror for Elastic EP Scaling, which only had a legacy MI300 job.
• models_distributed.yaml, quantization.yaml: the 2-GPU vision test on MI300 Distributed Models shard 2, and torchao 0.18.0 for the MI300 Quantization mirror ([Core] Update PyTorch to 2.14.0, torchvision to 0.29.0, triton to 3.8.0 (release channel) #52183 bumped NVIDIA and test-amd.yaml only).
• test-amd.yaml: regenerated as a copy of the test_areas AMD steps, with every step still optional.

test-amd.yaml and the test_areas mirrors have drifted apart. This PR moves into test_areas only the AMD coverage that test_areas doesn't already run on some AMD GPU, so no suite gets a second AMD platform. The regenerated file drops 84 legacy groups whose tests run in test_areas under a newer label or on another AMD GPU, including the five MRV2 groups #54823 removed. It also drops the gfx950 copies of Attention, MoE and Quantization Kernels, Quantization, Language Models (PPL), CUDAGraph and V1 Spec Decode, which run on MI300 or MI250 in test_areas (#57599 kept Quantization off MI355 on purpose), and gains 68 steps main CI runs but the legacy file never had. Rendered through both ci-infra generators, all 236 AMD jobs match on queue, commands, environment, timeout and shard count; main CI gains exactly the 3 new AMD steps and only the 3 edited steps change.

The added steps are copies of legacy jobs that passed every recent amd-ci main nightly they ran in (MoE Kernels Multi-GPU 8/8, Kimi K3 Multi-GPU 28/28, Elastic EP on MI300 28/28). On MI355 (gfx950), Kimi K3 Multi-GPU passed 10/10 and Elastic EP 14 of 16 (the other 2 failed when another process took the GPU memory). MI300 hardware wasn't available locally. #59677 makes the torchao and MI300 MoE changes too; whichever lands second drops them.

AI assistance (Claude Code) was used to draft this change. Every line was reviewed by the submitter.

---

Pull Request Checklist

• I used vLLM's /pr-checklist skill. (Mandatory for agents, optional for humans).

• AI assistance was used during the creation of this PR.

• Design Fit: Minimizes impact on core components, reuses existing functionality, and justifies added complexity.

• Testing and Validation: Validates the change and ensures any added tests are meaningful and reliable, with CI coverage or documented CI resource constraints and validation performed outside CI.

• Code Quality and Style: Keeps code and comments clear and concise, and updates relevant documentation and examples.

• Pull Request Contents: Includes a brief summary and relevant links, supports claims with evidence, explains root causes and implementation trade-offs, and follows the contributing guide.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @AndreasKaratzas.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork