[AMD][CI] Default the ROCm 7.2 PR gate to ROCm 7.2.4 Image (#35602)

This commit is contained in:
Bingxu Chen
2026-08-20 01:36:26 -07:00
committed by GitHub
parent 21c88f8625
commit f386e2a471
4 changed files with 18 additions and 16 deletions
+1
View File
@@ -128,6 +128,7 @@ jobs:
secrets: inherit
with:
ref: amd/aiter-ci
rocm_version: rocm724
run_all_tests: true
aiter_ref: ${{ needs.resolve-aiter.outputs.aiter_sha }}
continue_on_error: ${{ inputs.continue_on_error == '' && true || inputs.continue_on_error }}
+11 -11
View File
@@ -37,13 +37,13 @@ on:
- mi300
- mi325
rocm_version:
description: 'ROCm container variant (ROCm 7.2.0 by default)'
description: 'ROCm container variant (ROCm 7.2.4 by default)'
required: false
type: choice
default: rocm720
default: rocm724
options:
- rocm720
- rocm724
- rocm720
- rocm700
aiter_ref:
description: 'Override AITER commit (optional, leave empty to use Dockerfile default)'
@@ -73,10 +73,10 @@ on:
type: string
default: mi300
rocm_version:
description: 'ROCm container variant (ROCm 7.2.0 by default)'
description: 'ROCm container variant (ROCm 7.2.4 by default)'
required: false
type: string
default: rocm720
default: rocm724
aiter_ref:
description: 'Override AITER commit (optional, leave empty to use Dockerfile default)'
required: false
@@ -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 != 'rocm720' && 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 && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', 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') }}
@@ -158,7 +158,7 @@ jobs:
- name: Start CI container
# `inputs` is empty on pull_request events, so fall back explicitly to
# the ROCm 7.2 PR default. Reusable callers can request another variant.
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm720' }}
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm724' }}
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
@@ -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 != 'rocm720' && 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 && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', 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') }}
@@ -189,7 +189,7 @@ jobs:
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
- name: Start CI container
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm720' }}
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm724' }}
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
@@ -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 != 'rocm720' && 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 && inputs.rocm_version != 'rocm724' && format('-{0}', inputs.rocm_version) || '', 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') }}
@@ -219,7 +219,7 @@ jobs:
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
- name: Start CI container
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm720' }}
run: bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ inputs.rocm_version || 'rocm724' }}
env:
GITHUB_WORKSPACE: ${{ github.workspace }}
+5 -5
View File
@@ -20,10 +20,10 @@ on:
description: 'ROCm image version'
required: false
type: choice
default: rocm720
default: rocm724
options:
- rocm720
- rocm724
- rocm720
target_stage_select:
description: "Select a stage to run from dropdown (leave empty for auto-detect)"
required: false
@@ -94,7 +94,7 @@ on:
description: 'ROCm image version'
required: false
type: string
default: rocm720
default: rocm724
ref:
description: 'Git ref (branch, tag, or SHA) to test. If not provided, uses the default branch.'
required: false
@@ -140,7 +140,7 @@ permissions:
env:
AITER_COMMIT_OVERRIDE: ${{ inputs.aiter_ref }}
AMD_CI_IMAGE: ${{ inputs.amd_ci_image }}
ROCM_VERSION: ${{ inputs.rocm_version || 'rocm720' }}
ROCM_VERSION: ${{ inputs.rocm_version || 'rocm724' }}
DOCKERHUB_AMD_USERNAME: ${{ secrets.DOCKERHUB_AMD_USERNAME }}
DOCKERHUB_AMD_TOKEN: ${{ secrets.DOCKERHUB_AMD_TOKEN }}
@@ -250,7 +250,7 @@ jobs:
with:
ref: ${{ inputs.pr_head_sha || inputs.ref || '' }}
runner_arch: ${{ inputs.runner_arch || 'mi300' }}
rocm_version: ${{ inputs.rocm_version || 'rocm720' }}
rocm_version: ${{ inputs.rocm_version || 'rocm724' }}
aiter_ref: ${{ inputs.aiter_ref }}
amd_ci_image: ${{ inputs.amd_ci_image }}
continue_on_error: true
+1
View File
@@ -177,6 +177,7 @@ jobs:
uses: ./.github/workflows/pr-test-amd-rocm720.yml
with:
ref: ${{ needs.cut-release-branch.outputs.branch_name }}
rocm_version: rocm724
run_all_tests: true
secrets: inherit