[AMD][CI] Name the ROCm Image That Actually Ran in AMD Job Names (#35686)

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-24 01:07:25 -07:00
parent 5c4622341f
commit 7de80e566c
9 changed files with 327 additions and 88 deletions
@@ -129,26 +129,45 @@ jobs:
      - name: Parse workflow files
        id: parse
        run: |
+          set -euo pipefail
+
          # Parse the default ROCm 7.2 PR gate (exclude utility jobs)
-          # Excluded: call-gate, check-changes, finish, cancel, check-all-jobs
+          # Excluded: call-gate, check-changes, finish, cancel, check-all-jobs,
+          # plus the stage waits -- they run on ubuntu-latest, so the
+          # self-hosted-only snapshot has nothing for them. Keep the extra call:
+          # its delegated GPU jobs use the explicit caller name as their prefix.
+          # Job ids keep the -rocm720 lane suffix but display names no longer do,
+          # so strip it to get the prefix query_job_status.py matches against.
          pr_jobs=$(yq -r '.jobs | keys | .[]' .github/workflows/pr-test-amd-rocm720.yml | \
-            grep -v -E '^(call-gate|check-changes|pr-test-amd-rocm720-finish|cancel|check-all-jobs)$' | \
+            grep -v -E '^(call-gate|check-changes|pr-test-amd-rocm720-finish|cancel|check-all-jobs|wait-for-stage-[ab]-amd-rocm720)$' | \
+            sed 's/-rocm720$//' | \
            jq -R -s -c 'split("\n") | map(select(length > 0))')
          echo "pr_jobs=$pr_jobs" >> $GITHUB_OUTPUT
          echo "PR jobs: $pr_jobs"

          # Parse the default ROCm 7.2 nightly (exclude utility jobs)
          # Excluded: check-all-jobs
-          nightly_jobs=$(yq -r '.jobs | keys | .[]' .github/workflows/nightly-test-amd-rocm720.yml | \
-            grep -v -E '^(check-all-jobs)$' | \
+          # One nightly run covers every flavor, so ask for each by its full name:
+          # a stem-only prefix matches every flavor and merges them into a single
+          # row, hiding a regression that only reproduces on one image. Nightly
+          # names carry the runner too, so pair each job id with its runs-on.
+          # Keep in sync with the rocm_version matrix in the nightly workflow.
+          rocm_flavors="rocm724 rocm720"
+          nightly_jobs=$(yq -r '.jobs | to_entries[] | .key + " " + .value."runs-on"' .github/workflows/nightly-test-amd-rocm720.yml | \
+            grep -v -E '^check-all-jobs ' | \
+            while read -r id runner; do
+              stem=${id%-rocm720}
+              for f in $rocm_flavors; do echo "$stem ($f, $runner)"; done
+            done | \
            jq -R -s -c 'split("\n") | map(select(length > 0))')
          echo "nightly_jobs=$nightly_jobs" >> $GITHUB_OUTPUT
          echo "Nightly jobs: $nightly_jobs"

          # Parse the ROCm 7.0 rollback shadow (exclude utility jobs)
-          # Excluded: call-gate, check-changes, finish, cancel, check-all-jobs
+          # Excluded: call-gate, check-changes, finish, and the stage waits.
+          # Keep the reusable extra caller so its nested GPU jobs remain covered.
          pr_rocm700_jobs=$(yq -r '.jobs | keys | .[]' .github/workflows/pr-test-amd.yml | \
-            grep -v -E '^(call-gate|check-changes|pr-test-amd-finish|cancel|check-all-jobs)$' | \
+            grep -v -E '^(call-gate|check-changes|pr-test-amd-finish|wait-for-stage-[ab]-amd)$' | \
            jq -R -s -c 'split("\n") | map(select(length > 0))')
          echo "pr_rocm700_jobs=$pr_rocm700_jobs" >> $GITHUB_OUTPUT
          echo "PR ROCm 7.0 shadow jobs: $pr_rocm700_jobs"
@@ -161,6 +161,7 @@ jobs:
  # ==============================================================================

  nightly-test-1-gpu-unit-rocm720:
+    name: ${{ format('nightly-test-1-gpu-unit ({0}, linux-mi300-1gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -176,7 +177,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -195,6 +196,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-test-1-gpu-kernel-rocm720:
+    name: ${{ format('nightly-test-1-gpu-kernel ({0}, linux-mi300-1gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -210,7 +212,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -229,6 +231,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-test-1-gpu-mi35x-rocm720:
+    name: ${{ format('nightly-test-1-gpu-mi35x ({0}, linux-mi35x-gpu-1)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -244,7 +247,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -268,6 +271,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-2-gpu-rocm720:
+    name: ${{ format('nightly-accuracy-2-gpu ({0}, linux-mi300-2gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -283,7 +287,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -302,6 +306,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-2-gpu-mi35x-glm51-mxfp4-rocm720:
+    name: ${{ format('nightly-2-gpu-mi35x-glm51-mxfp4 ({0}, linux-mi35x-gpu-2)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -317,7 +322,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -340,6 +345,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-2-gpu-mi35x-deepseek-r1-mxfp4-tp2-rocm720:
+    name: ${{ format('nightly-2-gpu-mi35x-deepseek-r1-mxfp4-tp2 ({0}, linux-mi35x-gpu-2)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -355,7 +361,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -379,6 +385,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-2-gpu-vlm-rocm720:
+    name: ${{ format('nightly-accuracy-2-gpu-vlm ({0}, linux-mi300-2gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -394,7 +401,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -414,6 +421,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-2-gpu-text-rocm720:
+    name: ${{ format('nightly-perf-2-gpu-text ({0}, linux-mi300-2gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -429,7 +437,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -450,6 +458,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-2-gpu-vlm-rocm720:
+    name: ${{ format('nightly-perf-2-gpu-vlm ({0}, linux-mi300-2gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -465,7 +474,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -486,6 +495,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-4-gpu-rocm720:
+    name: ${{ format('nightly-4-gpu ({0}, linux-mi300-4gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -501,7 +511,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -526,6 +536,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-8-gpu-rocm720:
+    name: ${{ format('nightly-accuracy-8-gpu ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -541,7 +552,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -578,6 +589,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-8-gpu-mi35x-rocm720:
+    name: ${{ format('nightly-accuracy-8-gpu-mi35x ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -593,7 +605,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -634,6 +646,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-grok1-int4-rocm720:
+    name: ${{ format('nightly-8-gpu-grok1-int4 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -649,7 +662,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -683,6 +696,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-grok1-int4-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-grok1-int4 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -698,7 +712,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -739,6 +753,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-grok2-rocm720:
+    name: ${{ format('nightly-8-gpu-grok2 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -754,7 +769,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -788,6 +803,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-grok2-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-grok2 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -803,7 +819,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -844,6 +860,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-deepseek-v31-rocm720:
+    name: ${{ format('nightly-8-gpu-deepseek-v31 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -859,7 +876,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -894,6 +911,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v32-rocm720:
+    name: ${{ format('nightly-8-gpu-deepseek-v32 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -909,7 +927,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -942,6 +960,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v32-mtp-rocm720:
+    name: ${{ format('nightly-8-gpu-deepseek-v32-mtp ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -957,7 +976,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -990,6 +1009,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v3-kv-fp8-rocm720:
+    name: ${{ format('nightly-8-gpu-deepseek-v3-kv-fp8 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1005,7 +1025,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1031,6 +1051,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-8-gpu-mi35x-deepseek-v32-rocm720:
+    name: ${{ format('nightly-accuracy-8-gpu-mi35x-deepseek-v32 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1046,7 +1067,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1071,6 +1092,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp-rocm720:
+    name: ${{ format('nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1086,7 +1108,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1111,6 +1133,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-8-gpu-mi35x-deepseek-v32-basic-rocm720:
+    name: ${{ format('nightly-perf-8-gpu-mi35x-deepseek-v32-basic ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1126,7 +1149,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1151,6 +1174,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-8-gpu-mi35x-deepseek-v32-mtp-rocm720:
+    name: ${{ format('nightly-perf-8-gpu-mi35x-deepseek-v32-mtp ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1166,7 +1190,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1195,6 +1219,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-r1-mxfp4 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1210,7 +1235,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1246,6 +1271,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-tp4-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-r1-mxfp4-tp4 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1261,7 +1287,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1296,6 +1322,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1311,7 +1338,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1347,6 +1374,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1362,7 +1390,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1398,6 +1426,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-hicache-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-r1-hicache ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1413,7 +1442,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1440,6 +1469,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-v4-flash ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1455,7 +1485,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1495,6 +1525,7 @@ jobs:
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true

  nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-v4-pro ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1510,7 +1541,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1550,6 +1581,7 @@ jobs:
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true

  nightly-8-gpu-mi35x-deepseek-v4-pro-mtp-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-v4-pro-mtp ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1565,7 +1597,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1605,6 +1637,7 @@ jobs:
          echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true

  nightly-8-gpu-mi35x-deepseek-v4-pro-dspark-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-deepseek-v4-pro-dspark ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1620,7 +1653,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1654,6 +1687,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-kimi-k26-rocm720:
+    name: ${{ format('nightly-8-gpu-kimi-k26 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1669,7 +1703,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1695,6 +1729,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-kimi-k3-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-kimi-k3 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1710,7 +1745,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1758,6 +1793,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-qwen35-rocm720:
+    name: ${{ format('nightly-8-gpu-qwen35 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1773,7 +1809,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1810,6 +1846,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-qwen35-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-qwen35 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1825,7 +1862,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1861,6 +1898,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-qwen35-triton-dcp-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-qwen35-triton-dcp ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1876,7 +1914,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1903,6 +1941,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-qwen38-mxfp4-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-qwen38-mxfp4 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1918,7 +1957,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -1960,6 +1999,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-glm51-rocm720:
+    name: ${{ format('nightly-8-gpu-glm51 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -1975,7 +2015,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2013,6 +2053,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-glm52-fp8-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-glm52-fp8 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2028,7 +2069,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2071,6 +2112,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-glm5-mxfp4-rocm720:
+    name: ${{ format('nightly-8-gpu-mi35x-glm5-mxfp4 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2086,7 +2128,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2126,6 +2168,7 @@ jobs:
  # ==============================================================================

  nightly-4-gpu-mi35x-minimax-m25-rocm720:
+    name: ${{ format('nightly-4-gpu-mi35x-minimax-m25 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2141,7 +2184,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2169,6 +2212,7 @@ jobs:
  # ==============================================================================

  nightly-4-gpu-mi35x-minimax-m3-rocm720:
+    name: ${{ format('nightly-4-gpu-mi35x-minimax-m3 ({0}, linux-mi35x-gpu-8)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2184,7 +2228,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2211,6 +2255,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-minimax-m27-rocm720:
+    name: ${{ format('nightly-8-gpu-minimax-m27 ({0}, linux-mi300-8gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2226,7 +2271,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -2265,6 +2310,7 @@ jobs:
  # ==============================================================================

  nightly-1-gpu-zimage-turbo-rocm720:
+    name: ${{ format('nightly-1-gpu-zimage-turbo ({0}, linux-mi300-1gpu-sglang)', matrix.rocm_version) }}
    strategy:
      fail-fast: false
      matrix:
@@ -2280,7 +2326,7 @@ jobs:
      - name: Ensure VRAM is clear
        run: bash scripts/ci/amd/ensure_vram_clear.sh rocm

-      - name: Setup docker (ROCm 7.2)
+      - name: Setup docker (${{ matrix.rocm_version }})
        run: |
          touch github_summary.md
          bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
@@ -139,6 +139,7 @@ jobs:
  # ==============================================================================

  nightly-test-1-gpu-unit:
+    name: nightly-test-1-gpu-unit (rocm700, linux-mi300-1gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-test-1-gpu-unit,'))
    runs-on: linux-mi300-1gpu-sglang
    steps:
@@ -170,6 +171,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-test-1-gpu-kernel:
+    name: nightly-test-1-gpu-kernel (rocm700, linux-mi300-1gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-test-1-gpu-kernel,'))
    runs-on: linux-mi300-1gpu-sglang
    steps:
@@ -201,6 +203,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-test-1-gpu-mi35x:
+    name: nightly-test-1-gpu-mi35x (rocm700, linux-mi35x-gpu-1)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-test-1-gpu-mi35x,'))
    runs-on: linux-mi35x-gpu-1
    steps:
@@ -239,6 +242,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-2-gpu:
+    name: nightly-accuracy-2-gpu (rocm700, linux-mi300-2gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-2-gpu,'))
    runs-on: linux-mi300-2gpu-sglang
    steps:
@@ -270,6 +274,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-2-gpu-mi35x-glm51-mxfp4:
+    name: nightly-2-gpu-mi35x-glm51-mxfp4 (rocm700, linux-mi35x-gpu-2)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-2-gpu-mi35x-glm51-mxfp4,'))
    runs-on: linux-mi35x-gpu-2
    steps:
@@ -304,6 +309,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-2-gpu-mi35x-deepseek-r1-mxfp4-tp2:
+    name: nightly-2-gpu-mi35x-deepseek-r1-mxfp4-tp2 (rocm700, linux-mi35x-gpu-2)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-2-gpu-mi35x-deepseek-r1-mxfp4-tp2,'))
    runs-on: linux-mi35x-gpu-2
    steps:
@@ -339,6 +345,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-2-gpu-vlm:
+    name: nightly-accuracy-2-gpu-vlm (rocm700, linux-mi300-2gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-2-gpu-vlm,'))
    runs-on: linux-mi300-2gpu-sglang
    steps:
@@ -371,6 +378,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-2-gpu-text:
+    name: nightly-perf-2-gpu-text (rocm700, linux-mi300-2gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-2-gpu-text,'))
    runs-on: linux-mi300-2gpu-sglang
    steps:
@@ -404,6 +412,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-2-gpu-vlm:
+    name: nightly-perf-2-gpu-vlm (rocm700, linux-mi300-2gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-2-gpu-vlm,'))
    runs-on: linux-mi300-2gpu-sglang
    steps:
@@ -437,6 +446,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-4-gpu:
+    name: nightly-4-gpu (rocm700, linux-mi300-4gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-4-gpu,'))
    runs-on: linux-mi300-4gpu-sglang
    steps:
@@ -473,6 +483,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-8-gpu:
+    name: nightly-accuracy-8-gpu (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -504,6 +515,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-8-gpu-mi35x:
+    name: nightly-accuracy-8-gpu-mi35x (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -542,6 +554,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-grok1-int4:
+    name: nightly-8-gpu-grok1-int4 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-grok1-int4,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -587,6 +600,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-grok1-int4:
+    name: nightly-8-gpu-mi35x-grok1-int4 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-grok1-int4,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -639,6 +653,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-grok2:
+    name: nightly-8-gpu-grok2 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-grok2,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -684,6 +699,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-grok2:
+    name: nightly-8-gpu-mi35x-grok2 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-grok2,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -736,6 +752,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-deepseek-v31:
+    name: nightly-8-gpu-deepseek-v31 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v31,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -782,6 +799,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v32:
+    name: nightly-8-gpu-deepseek-v32 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v32,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -826,6 +844,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v32-mtp:
+    name: nightly-8-gpu-deepseek-v32-mtp (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v32-mtp,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -870,6 +889,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-deepseek-v3-kv-fp8:
+    name: nightly-8-gpu-deepseek-v3-kv-fp8 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-deepseek-v3-kv-fp8,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -907,6 +927,7 @@ jobs:
  # ==============================================================================

  nightly-accuracy-8-gpu-mi35x-deepseek-v32:
+    name: nightly-accuracy-8-gpu-mi35x-deepseek-v32 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x-deepseek-v32,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -943,6 +964,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp:
+    name: nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-accuracy-8-gpu-mi35x-deepseek-v32-mtp,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -979,6 +1001,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-8-gpu-mi35x-deepseek-v32-basic:
+    name: nightly-perf-8-gpu-mi35x-deepseek-v32-basic (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-8-gpu-mi35x-deepseek-v32-basic,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1015,6 +1038,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-perf-8-gpu-mi35x-deepseek-v32-mtp:
+    name: nightly-perf-8-gpu-mi35x-deepseek-v32-mtp (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-perf-8-gpu-mi35x-deepseek-v32-mtp,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1055,6 +1079,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4:
+    name: nightly-8-gpu-mi35x-deepseek-r1-mxfp4 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1102,6 +1127,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-tp4:
+    name: nightly-8-gpu-mi35x-deepseek-r1-mxfp4-tp4 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-tp4,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1148,6 +1174,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8:
+    name: nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-kv-fp8,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1195,6 +1222,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion:
+    name: nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-mxfp4-ar-fusion,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1242,6 +1270,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-deepseek-r1-hicache:
+    name: nightly-8-gpu-mi35x-deepseek-r1-hicache (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-deepseek-r1-hicache,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1280,6 +1309,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-kimi-k26:
+    name: nightly-8-gpu-kimi-k26 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-kimi-k26,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -1313,6 +1343,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-kimi-k26:
+    name: nightly-8-gpu-mi35x-kimi-k26 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-kimi-k26,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1353,6 +1384,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-qwen3-235b:
+    name: nightly-8-gpu-qwen3-235b (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-qwen3-235b,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -1386,6 +1418,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-qwen3-235b-mxfp4:
+    name: nightly-8-gpu-mi35x-qwen3-235b-mxfp4 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-qwen3-235b-mxfp4,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1425,6 +1458,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-qwen35:
+    name: nightly-8-gpu-qwen35 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-qwen35,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -1473,6 +1507,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-qwen35:
+    name: nightly-8-gpu-mi35x-qwen35 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-qwen35,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1520,6 +1555,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-qwen35-triton-dcp:
+    name: nightly-8-gpu-mi35x-qwen35-triton-dcp (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-qwen35-triton-dcp,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1558,6 +1594,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-glm51:
+    name: nightly-8-gpu-glm51 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-glm51,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -1603,6 +1640,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  nightly-8-gpu-mi35x-glm51:
+    name: nightly-8-gpu-mi35x-glm51 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-glm51,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1652,6 +1690,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-mi35x-glm5-mxfp4:
+    name: nightly-8-gpu-mi35x-glm5-mxfp4 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-mi35x-glm5-mxfp4,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1703,6 +1742,7 @@ jobs:
  # ==============================================================================

  nightly-4-gpu-mi35x-minimax-m25:
+    name: nightly-4-gpu-mi35x-minimax-m25 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-4-gpu-mi35x-minimax-m25,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1742,6 +1782,7 @@ jobs:
  # ==============================================================================

  nightly-4-gpu-mi35x-minimax-m3:
+    name: nightly-4-gpu-mi35x-minimax-m3 (rocm700, linux-mi35x-gpu-8)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-4-gpu-mi35x-minimax-m3,'))
    runs-on: linux-mi35x-gpu-8
    steps:
@@ -1780,6 +1821,7 @@ jobs:
  # ==============================================================================

  nightly-8-gpu-minimax-m27:
+    name: nightly-8-gpu-minimax-m27 (rocm700, linux-mi300-8gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-8-gpu-minimax-m27,'))
    runs-on: linux-mi300-8gpu-sglang
    steps:
@@ -1830,6 +1872,7 @@ jobs:
  # ==============================================================================

  nightly-1-gpu-zimage-turbo:
+    name: nightly-1-gpu-zimage-turbo (rocm700, linux-mi300-1gpu-sglang)
    if: (github.repository == 'sgl-project/sglang' || github.event_name == 'pull_request') && (!(inputs.job_filter || inputs.job_select) || (inputs.job_filter || inputs.job_select) == 'all' || contains(format(',{0},', inputs.job_filter || inputs.job_select), ',nightly-1-gpu-zimage-turbo,'))
    runs-on: linux-mi300-1gpu-sglang
    steps:
@@ -142,7 +142,7 @@ jobs:
  # setup across scarce AMD GPUs to shave only a couple minutes of test time,
  # so one GPU running the whole suite sequentially is the better trade.
  extra-a-test-1-gpu-small-amd:
-    name: ${{ format('extra-a-test-1-gpu-small-amd{0} (linux-{1}-1gpu-sglang)', inputs.rocm_version && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('extra-a-test-1-gpu-small-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [call-gate]
    if: ${{ !cancelled() && needs.call-gate.result == 'success' }}
    runs-on: ${{ format('linux-{0}-1gpu-sglang', inputs.runner_arch || 'mi300') }}
@@ -175,7 +175,7 @@ jobs:
  # pool as small (AMD GPUs carry enough VRAM that "large" here is a CUDA
  # memory-tier label, not a separate AMD runner pool).
  extra-a-test-1-gpu-large-amd:
-    name: ${{ format('extra-a-test-1-gpu-large-amd{0} (linux-{1}-1gpu-sglang)', inputs.rocm_version && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('extra-a-test-1-gpu-large-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [call-gate]
    if: ${{ !cancelled() && needs.call-gate.result == 'success' }}
    runs-on: ${{ format('linux-{0}-1gpu-sglang', inputs.runner_arch || 'mi300') }}
@@ -205,7 +205,7 @@ jobs:
  # Multi-GPU TP / PP / PD mock-model + kv_canary e2e tests. Mirrors CUDA's
  # extra-a 2-gpu-large stage; runs on the 2-GPU AMD pool.
  extra-a-test-2-gpu-large-amd:
-    name: ${{ format('extra-a-test-2-gpu-large-amd{0} (linux-{1}-2gpu-sglang)', inputs.rocm_version && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('extra-a-test-2-gpu-large-amd ({0}, linux-{1}-2gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [call-gate]
    if: ${{ !cancelled() && needs.call-gate.result == 'success' }}
    runs-on: ${{ format('linux-{0}-2gpu-sglang', inputs.runner_arch || 'mi300') }}
@@ -242,6 +242,7 @@ jobs:
  # Not added to `pr-test-amd-rocm720-finish` so the base AMD gate never depends on
  # the opt-in extra suite.
  call-pr-test-amd-extra-rocm720:
+    name: call-pr-test-amd-extra
    if: |
      (github.event_name == 'schedule' || inputs.run_all_tests == true) &&
      !(inputs.target_stage || inputs.target_stage_select)
@@ -257,7 +258,7 @@ jobs:

  # =============================================== sgl-kernel ====================================================
  sgl-kernel-unit-test-amd-rocm720:
-    name: ${{ format('sgl-kernel-unit-test-amd-rocm720 (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('sgl-kernel-unit-test-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -315,7 +316,7 @@ jobs:
          exit $failures

  sgl-kernel-unit-test-2-gpu-amd-rocm720:
-    name: ${{ format('sgl-kernel-unit-test-2-gpu-amd-rocm720 (linux-{0}-2gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('sgl-kernel-unit-test-2-gpu-amd ({0}, linux-{1}-2gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -356,7 +357,7 @@ jobs:
  # =============================================== primary ====================================================

  stage-a-test-1-gpu-small-amd-rocm720:
-    name: ${{ format('stage-a-test-1-gpu-small-amd-rocm720 (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('stage-a-test-1-gpu-small-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -393,7 +394,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-a-test-1-gpu-small-amd ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  jit-kernel-unit-test-amd-rocm720:
-    name: ${{ format('jit-kernel-unit-test-amd-rocm720 (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('jit-kernel-unit-test-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -430,7 +431,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite jit-kernel-unit-test-amd ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  jit-kernel-benchmark-test-amd-rocm720:
-    name: ${{ format('jit-kernel-benchmark-test-amd-rocm720 (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('jit-kernel-benchmark-test-amd ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -472,6 +473,7 @@ jobs:
  # For scheduled runs: wait jobs are skipped, enabling parallel execution of all stages.

  wait-for-stage-a-amd-rocm720:
+    name: wait-for-stage-a-amd
    needs: [check-changes, call-gate]
    if: |
      always() &&
@@ -488,12 +490,12 @@ jobs:
      - uses: ./.github/actions/wait-for-jobs
        id: wait
        with:
-          stage-name: stage-a-amd-rocm720
-          jobs: '[{"prefix": "stage-a-test-1-gpu-small-amd-rocm720", "expected_count": 1}]'
+          stage-name: stage-a-amd
+          jobs: '[{"prefix": "stage-a-test-1-gpu-small-amd", "expected_count": 1}]'
          max-wait-minutes: '240'

  stage-b-test-1-gpu-small-amd-rocm720:
-    name: ${{ format('stage-b-test-1-gpu-small-amd-rocm720 (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-1-gpu-small-amd ({0}, linux-{1}-1gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -534,7 +536,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 14 --timeout-per-file 2400 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-small-amd-nondeterministic-rocm720:
-    name: ${{ format('stage-b-test-1-gpu-small-amd-nondeterministic-rocm720 (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('stage-b-test-1-gpu-small-amd-nondeterministic ({0}, linux-{1}-1gpu-sglang)', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -570,6 +572,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd-nondeterministic --timeout-per-file 3600 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-small-amd-mi35x-rocm720:
+    name: ${{ format('stage-b-test-1-gpu-small-amd-mi35x ({0}, {1})', inputs.rocm_version || 'rocm724', matrix.runner) }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -609,7 +612,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd-mi35x ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-large-amd-rocm720:
-    name: ${{ format('stage-b-test-1-gpu-large-amd-rocm720 (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-1-gpu-large-amd ({0}, linux-{1}-1gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -650,7 +653,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-large-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 2700 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-2-gpu-large-amd-rocm720:
-    name: ${{ format('stage-b-test-2-gpu-large-amd-rocm720 (linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-2-gpu-large-amd ({0}, linux-{1}-2gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -691,7 +694,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-2-gpu-large-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 2 --timeout-per-file ${{ (inputs.runner_arch || 'mi300') == 'mi300' && 5400 || 2700 }} ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  multimodal-gen-test-1-gpu-amd-rocm720:
-    name: ${{ format('multimodal-gen-test-1-gpu-amd-rocm720 (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('multimodal-gen-test-1-gpu-amd ({0}, linux-{1}-1gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -840,7 +843,7 @@ jobs:
          retention-days: 7

  multimodal-gen-test-2-gpu-amd-rocm720:
-    name: ${{ format('multimodal-gen-test-2-gpu-amd-rocm720 (linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('multimodal-gen-test-2-gpu-amd ({0}, linux-{1}-2gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -975,6 +978,7 @@ jobs:
          retention-days: 7

  wait-for-stage-b-amd-rocm720:
+    name: wait-for-stage-b-amd
    needs: [check-changes, call-gate, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -992,17 +996,17 @@ jobs:
      - uses: ./.github/actions/wait-for-jobs
        id: wait
        with:
-          stage-name: stage-b-amd-rocm720
+          stage-name: stage-b-amd
          jobs: |
            [
-              {"prefix": "stage-b-test-1-gpu-small-amd-rocm720", "expected_count": 14},
-              {"prefix": "stage-b-test-1-gpu-large-amd-rocm720", "expected_count": 3},
-              {"prefix": "stage-b-test-2-gpu-large-amd-rocm720", "expected_count": 2}
+              {"prefix": "stage-b-test-1-gpu-small-amd", "expected_count": 14},
+              {"prefix": "stage-b-test-1-gpu-large-amd", "expected_count": 3},
+              {"prefix": "stage-b-test-2-gpu-large-amd", "expected_count": 2}
            ]
          max-wait-minutes: '480'

  stage-c-test-4-gpu-amd-rocm720:
-    name: ${{ format('stage-c-test-4-gpu-amd-rocm720 (linux-{0}-4gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-c-test-4-gpu-amd ({0}, linux-{1}-4gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd-rocm720]
    if: |
      always() &&
@@ -1058,7 +1062,7 @@ jobs:
              ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-c-test-large-8-gpu-amd-rocm720:
-    name: ${{ format('stage-c-test-large-8-gpu-amd-rocm720 (linux-{0}-8gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-c-test-large-8-gpu-amd ({0}, linux-{1}-8gpu-sglang, {2})', inputs.rocm_version || 'rocm724', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd-rocm720]
    if: |
      always() &&
@@ -1107,6 +1111,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-c-test-large-8-gpu-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 4 --timeout-per-file 5400 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-c-test-large-8-gpu-amd-mi35x-rocm720:
+    name: ${{ format('stage-c-test-large-8-gpu-amd-mi35x ({0}, {1}, {2})', inputs.rocm_version || 'rocm724', matrix.runner, matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd-rocm720]
    if: |
      always() &&
@@ -1150,6 +1155,7 @@ jobs:
  # GSM8K accuracy on the nightly dsv4 suites, ~20min each. Gated as stage-C so
  # every PR that touches the main package covers DeepSeek-V4.
  stage-c-dsv4-flash-fp4-fp8-amd-mi35x-rocm720:
+    name: ${{ format('stage-c-dsv4-flash-fp4-fp8-amd-mi35x ({0}, {1})', inputs.rocm_version || 'rocm724', matrix.runner) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd-rocm720]
    if: |
      always() &&
@@ -1204,6 +1210,7 @@ jobs:
          exit ${TEST_EXIT_CODE:-0}

  stage-c-dsv4-pro-fp4-amd-mi35x-rocm720:
+    name: ${{ format('stage-c-dsv4-pro-fp4-amd-mi35x ({0}, {1})', inputs.rocm_version || 'rocm724', matrix.runner) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd-rocm720]
    if: |
      always() &&
@@ -1259,6 +1266,7 @@ jobs:

  # =============================================== Disaggregation ====================================================
  stage-b-test-large-8-gpu-mi35x-disaggregation-amd-rocm720:
+    name: ${{ format('stage-b-test-large-8-gpu-mi35x-disaggregation-amd ({0}, {1})', inputs.rocm_version || 'rocm724', matrix.runner) }}
    needs: [check-changes, wait-for-stage-a-amd-rocm720]
    if: |
      always() &&
@@ -1392,6 +1400,7 @@ jobs:
            -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-large-8-gpu-mi35x-disaggregation-amd --timeout-per-file 1800 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  pr-test-amd-rocm720-finish:
+    name: pr-test-amd-finish
    needs:
      [
        call-gate,
@@ -226,7 +226,7 @@ jobs:

  # =============================================== sgl-kernel ====================================================
  sgl-kernel-unit-test-amd:
-    name: ${{ format('sgl-kernel-unit-test-amd (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('sgl-kernel-unit-test-amd (rocm700, linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -284,7 +284,7 @@ jobs:
          exit $failures

  sgl-kernel-unit-test-2-gpu-amd:
-    name: ${{ format('sgl-kernel-unit-test-2-gpu-amd (linux-{0}-2gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('sgl-kernel-unit-test-2-gpu-amd (rocm700, linux-{0}-2gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -325,7 +325,7 @@ jobs:
  # =============================================== primary ====================================================

  stage-a-test-1-gpu-small-amd:
-    name: ${{ format('stage-a-test-1-gpu-small-amd (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('stage-a-test-1-gpu-small-amd (rocm700, linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -362,7 +362,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-a-test-1-gpu-small-amd ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  jit-kernel-unit-test-amd:
-    name: ${{ format('jit-kernel-unit-test-amd (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('jit-kernel-unit-test-amd (rocm700, linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -399,7 +399,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite jit-kernel-unit-test-amd ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  jit-kernel-benchmark-test-amd:
-    name: ${{ format('jit-kernel-benchmark-test-amd (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('jit-kernel-benchmark-test-amd (rocm700, linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -462,7 +462,7 @@ jobs:
          max-wait-minutes: '240'

  stage-b-test-1-gpu-small-amd:
-    name: ${{ format('stage-b-test-1-gpu-small-amd (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-1-gpu-small-amd (rocm700, linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -502,7 +502,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 14 --timeout-per-file 2400 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-small-amd-nondeterministic:
-    name: ${{ format('stage-b-test-1-gpu-small-amd-nondeterministic (linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
+    name: ${{ format('stage-b-test-1-gpu-small-amd-nondeterministic (rocm700, linux-{0}-1gpu-sglang)', inputs.runner_arch || 'mi300') }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -538,6 +538,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd-nondeterministic --timeout-per-file 3600 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-small-amd-mi35x:
+    name: ${{ format('stage-b-test-1-gpu-small-amd-mi35x (rocm700, {0})', matrix.runner) }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -577,7 +578,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-small-amd-mi35x ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-1-gpu-large-amd:
-    name: ${{ format('stage-b-test-1-gpu-large-amd (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-1-gpu-large-amd (rocm700, linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -617,7 +618,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-1-gpu-large-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 3 --timeout-per-file 2700 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-b-test-2-gpu-large-amd:
-    name: ${{ format('stage-b-test-2-gpu-large-amd (linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-b-test-2-gpu-large-amd (rocm700, linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -657,7 +658,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-b-test-2-gpu-large-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 2 --timeout-per-file ${{ (inputs.runner_arch || 'mi300') == 'mi300' && 5400 || 2700 }} ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  multimodal-gen-test-1-gpu-amd:
-    name: ${{ format('multimodal-gen-test-1-gpu-amd (linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('multimodal-gen-test-1-gpu-amd (rocm700, linux-{0}-1gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -805,7 +806,7 @@ jobs:
          retention-days: 7

  multimodal-gen-test-2-gpu-amd:
-    name: ${{ format('multimodal-gen-test-2-gpu-amd (linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('multimodal-gen-test-2-gpu-amd (rocm700, linux-{0}-2gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate]
    if: |
      always() && !cancelled() &&
@@ -966,7 +967,7 @@ jobs:
          max-wait-minutes: '480'

  stage-c-test-4-gpu-amd:
-    name: ${{ format('stage-c-test-4-gpu-amd (linux-{0}-4gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-c-test-4-gpu-amd (rocm700, linux-{0}-4gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd]
    if: |
      always() &&
@@ -1022,7 +1023,7 @@ jobs:
              ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-c-test-large-8-gpu-amd:
-    name: ${{ format('stage-c-test-large-8-gpu-amd (linux-{0}-8gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
+    name: ${{ format('stage-c-test-large-8-gpu-amd (rocm700, linux-{0}-8gpu-sglang, {1})', inputs.runner_arch || 'mi300', matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd]
    if: |
      always() &&
@@ -1070,6 +1071,7 @@ jobs:
          bash scripts/ci/amd/amd_ci_exec.sh -w "/sglang-checkout/test" python3 run_suite.py --hw amd --suite stage-c-test-large-8-gpu-amd --auto-partition-id ${{ matrix.part }} --auto-partition-size 4 --timeout-per-file 5400 ${{ needs.check-changes.outputs.continue_on_error == 'true' && '--continue-on-error' || '' }}

  stage-c-test-large-8-gpu-amd-mi35x:
+    name: ${{ format('stage-c-test-large-8-gpu-amd-mi35x (rocm700, {0}, {1})', matrix.runner, matrix.part) }}
    needs: [check-changes, call-gate, wait-for-stage-b-amd]
    if: |
      always() &&
@@ -1111,6 +1113,7 @@ jobs:

  # =============================================== Disaggregation ====================================================
  stage-b-test-large-8-gpu-mi35x-disaggregation-amd:
+    name: ${{ format('stage-b-test-large-8-gpu-mi35x-disaggregation-amd (rocm700, {0})', matrix.runner) }}
    needs: [check-changes, wait-for-stage-a-amd]
    if: |
      always() &&
@@ -6,7 +6,9 @@ Usage:
    # Per-job reports (original mode)
    python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x"
    python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x" --hours 48
-    python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x-rocm720" --workflow "pr-test-amd-rocm720.yml" --input-data-file actions-job-snapshot.json --summary
+    # A stem matches every ROCm flavor; use the full display name to pin one
+    python scripts/ci/utils/query_job_status.py --job "stage-c-test-large-8-gpu-amd-mi35x" --workflow "pr-test-amd-rocm720.yml" --input-data-file actions-job-snapshot.json --summary
+    python scripts/ci/utils/query_job_status.py --job "nightly-test-1-gpu-unit (rocm724, linux-mi300-1gpu-sglang)" --workflow "nightly-test-amd-rocm720.yml" --input-data-file actions-job-snapshot.json --summary

    # Runner fleet report (cross-workflow runner analytics)
    python scripts/ci/utils/query_job_status.py --runner-report --workflow "pr-test-amd-rocm720.yml,nightly-test-amd-rocm720.yml" --hours 24
@@ -27,6 +27,51 @@ from typing import Dict, List, Optional, Tuple
 import requests


+def _filter_legacy_amd_job_rows(job_data: Dict[str, Dict]) -> Dict[str, Dict]:
+    """Drop pre-cutover AMD names without changing the shared analyzer."""
+
+    filtered = {}
+    for full_name, data in job_data.items():
+        # This caller was renamed by the AMD job-name cutover. Other outer
+        # callers, including AITER's *-rocm720 callers, are still current.
+        name_parts = full_name.split(" / ")
+        if "call-pr-test-amd-extra-rocm720" in name_parts[:-1]:
+            continue
+
+        leaf_name = name_parts[-1]
+        if leaf_name.startswith(("wait-for-stage-a-amd", "wait-for-stage-b-amd")):
+            continue
+        if leaf_name in {
+            "call-gate",
+            "call-pr-test-amd-extra",
+            "check-all-jobs",
+            "check-changes",
+            "pr-gate",
+            "pr-test-amd-extra-finish",
+            "pr-test-amd-finish",
+            "pr-test-amd-rocm720-finish",
+        }:
+            continue
+
+        # Old display-name stems ended in -rocm<digits>. Their version was not
+        # reliable enough to map forward: a -rocm720 PR job could run rocm724.
+        stem = leaf_name.split(" (", 1)[0]
+        _, separator, version = stem.rpartition("-rocm")
+        if separator and version.isdigit():
+            continue
+
+        # An intermediate nightly schema showed only the ROCm flavor. Current
+        # names always include both the flavor and runner inside parentheses.
+        if leaf_name.endswith(")") and " (" in leaf_name:
+            details = leaf_name.rsplit(" (", 1)[1][:-1]
+            if details.startswith("rocm") and details[4:].isdigit():
+                continue
+
+        filtered[full_name] = data
+
+    return filtered
+
+
 class SGLangFailuresAnalyzer:
    """Analyzes consecutive failures in GitHub Actions workflows."""

@@ -2648,6 +2693,21 @@ def main():
            else ({}, {})
        )

+        # AMD renamed its display names to carry the actual ROCm flavor. Reset
+        # only AMD history at that boundary so legacy failures cannot remain
+        # "current" under names that no longer exist. Other platforms continue
+        # to use the unmodified shared analyzer results. General AMD reports
+        # intentionally omit old-schema rows from branches that have not moved
+        # to the new names; their ROCm flavor cannot be mapped reliably.
+        pr_test_amd_scheduled_data = _filter_legacy_amd_job_rows(
+            pr_test_amd_scheduled_data
+        )
+        nightly_amd_scheduled_data = _filter_legacy_amd_job_rows(
+            nightly_amd_scheduled_data
+        )
+        pr_test_amd_general_data = _filter_legacy_amd_job_rows(pr_test_amd_general_data)
+        nightly_amd_general_data = _filter_legacy_amd_job_rows(nightly_amd_general_data)
+
        # Analyze runner health and consecutive failures on all runs
        (
            runner_stats,
@@ -0,0 +1,57 @@
+"""AMD job-name cutover tests.
+
+Run with:
+    python -m unittest discover -s scripts/ci_monitor -p 'test_ci_failures_analysis.py'
+"""
+
+import os
+import sys
+import unittest
+
+sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
+from ci_failures_analysis import _filter_legacy_amd_job_rows  # noqa: E402
+
+
+class TestFilterLegacyAmdJobRows(unittest.TestCase):
+    def test_drops_legacy_names_and_nested_utilities(self):
+        rows = {
+            "stage-b-test-1-gpu-small-amd-rocm720 (linux-mi300-1gpu-sglang, 0)": {},
+            "nightly-accuracy-2-gpu-rocm720 (rocm724)": {},
+            "nightly-accuracy-2-gpu-rocm724": {},
+            "nightly-test-1-gpu-unit (rocm724)": {},
+            "call-pr-test-amd-rocm720 / call-pr-test-amd-extra-rocm720 / extra-a-test-1-gpu-small-amd (linux-mi300-1gpu-sglang)": {},
+            "wait-for-stage-a-amd": {},
+            "call-pr-test-amd-extra / pr-test-amd-extra-finish": {},
+            "call-pr-test-amd-extra / call-gate / pr-gate": {},
+        }
+
+        self.assertEqual(_filter_legacy_amd_job_rows(rows), {})
+
+    def test_keeps_current_flavors_and_nested_callers_separate(self):
+        new_success = {"current_streak": 0}
+        rows = {
+            "stage-b-test-1-gpu-small-amd (rocm724, linux-mi300-1gpu-sglang, 0)": new_success,
+            "nightly-accuracy-2-gpu (rocm720, linux-mi300-2gpu-sglang)": {
+                "current_streak": 1
+            },
+            "call-pr-test-amd-rocm720 / stage-c-test-4-gpu-amd (rocm724, linux-mi300-4gpu-sglang, 0)": {
+                "current_streak": 0
+            },
+            "call-pr-test-amd-rocm720 / call-pr-test-amd-extra / extra-a-test-1-gpu-small-amd (rocm724, linux-mi300-1gpu-sglang)": {
+                "current_streak": 0
+            },
+        }
+
+        filtered = _filter_legacy_amd_job_rows(rows)
+
+        self.assertEqual(set(filtered), set(rows))
+        self.assertIs(
+            filtered[
+                "stage-b-test-1-gpu-small-amd (rocm724, linux-mi300-1gpu-sglang, 0)"
+            ],
+            new_success,
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()