Revert "CI: Temporarily use previous vllm nightly image"

[gyohuangxin]

Reverts #3043

[github-actions]

🏷️ CI Guide

Runs automatically on every PR:

• ✅ Pre-checks (submodule verification, code formatting)
• ✅ Aiter op tests (gfx942 + gfx950)
• ✅ Triton tests on MI35X (only when aiter/ops/triton/** or related paths are changed)

Extended tests (opt-in via labels):

Label	Tests
ci:triton-300x	Run an additional Triton test job on MI300X in PRs; main branch always runs both MI35X and MI300X
ci:sglang	SGLang integration tests: DeepSeek-R1-MXFP4 accuracy, Qwen 3.5 accuracy
ci:atom	ATOM benchmark: DeepSeek-R1-0528, GPT-OSS-120B
ci:vllm	vLLM benchmark: GPT-OSS-120B, DeepSeek-R1-0528, Kimi-K2.5
ci:all	All of the above

Add labels via the sidebar or gh pr edit 3102 --add-label <label>

[Copilot]

Pull request overview

This PR reverts #3043 by changing the vLLM benchmark workflow’s Docker base image back from a pinned nightly digest to the floating rocm/vllm-dev:nightly tag.

Changes:

• Update .github/workflows/vllm_benchmark.yaml to use rocm/vllm-dev:nightly instead of a @sha256:-pinned image digest.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.