[AMD][CI] Add MiniMax-M3-MXFP8 MI35x nightly perf benchmark (#36142)
This commit is contained in:
@@ -96,9 +96,7 @@ on:
|
||||
- nightly-8-gpu-mi35x-glm52-fp8-rocm720
|
||||
# 8-GPU GLM-5-MXFP4 (MI35x only)
|
||||
- nightly-8-gpu-mi35x-glm5-mxfp4-rocm720
|
||||
# 4-GPU MiniMax-M2.5 (MI35x)
|
||||
- nightly-4-gpu-mi35x-minimax-m25-rocm720
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x)
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x accuracy + performance)
|
||||
- nightly-4-gpu-mi35x-minimax-m3-rocm720
|
||||
# 8-GPU MiniMax-M2.7 (MI30x only)
|
||||
- nightly-8-gpu-minimax-m27-rocm720
|
||||
@@ -2164,51 +2162,7 @@ jobs:
|
||||
exit ${TEST_EXIT_CODE:-0}
|
||||
|
||||
# ==============================================================================
|
||||
# 4-GPU MiniMax-M2.5 (MI35x)
|
||||
# ==============================================================================
|
||||
|
||||
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:
|
||||
rocm_version: ${{ fromJson(inputs.rocm_version && inputs.rocm_version != 'all' && format('["{0}"]', inputs.rocm_version) || '["rocm724", "rocm720"]') }}
|
||||
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-rocm720,'))
|
||||
runs-on: linux-mi35x-gpu-8
|
||||
steps:
|
||||
- name: Checkout code
|
||||
uses: actions/checkout@v4
|
||||
with:
|
||||
ref: ${{ inputs.ref || github.sha }}
|
||||
|
||||
- name: Ensure VRAM is clear
|
||||
run: bash scripts/ci/amd/ensure_vram_clear.sh rocm
|
||||
|
||||
- name: Setup docker (${{ matrix.rocm_version }})
|
||||
run: |
|
||||
touch github_summary.md
|
||||
bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version ${{ matrix.rocm_version }}
|
||||
env:
|
||||
GITHUB_WORKSPACE: ${{ github.workspace }}
|
||||
ENABLE_CACHE_HOST: "1"
|
||||
|
||||
- name: Install dependencies
|
||||
run: bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps
|
||||
|
||||
- name: Accuracy Test MI35x ROCm 7.2 (4-GPU MiniMax-M2.5)
|
||||
timeout-minutes: 120
|
||||
run: |
|
||||
> github_summary.md
|
||||
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
||||
-e SGLANG_USE_AITER=1 \
|
||||
-e SGLANG_USE_AITER_UNIFIED_ATTN=1 \
|
||||
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
||||
python3 run_suite.py --hw amd --suite nightly-amd-4-gpu-mi35x-minimax-m25-tp4 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
||||
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
||||
exit ${TEST_EXIT_CODE:-0}
|
||||
|
||||
# ==============================================================================
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x)
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x accuracy + performance)
|
||||
# ==============================================================================
|
||||
|
||||
nightly-4-gpu-mi35x-minimax-m3-rocm720:
|
||||
@@ -2250,6 +2204,20 @@ jobs:
|
||||
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
||||
exit ${TEST_EXIT_CODE:-0}
|
||||
|
||||
# Run perf only after correctness passes. Keeping both steps in this job
|
||||
# reuses the cached checkpoint and one scarce MI35x slot.
|
||||
- name: Performance Test MI35x ROCm 7.2 (4-GPU MiniMax-M3 MXFP8)
|
||||
timeout-minutes: 120
|
||||
continue-on-error: true # Perf test failure doesn't fail the job if accuracy passed
|
||||
run: |
|
||||
> github_summary.md
|
||||
bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \
|
||||
-e SGLANG_USE_AITER=1 \
|
||||
-e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \
|
||||
python3 run_suite.py --hw amd --suite nightly-perf-4-gpu-mi35x-minimax-m3 --nightly --timeout-per-file 5400 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} || TEST_EXIT_CODE=$?
|
||||
echo "$(<github_summary.md )" >> $GITHUB_STEP_SUMMARY || true
|
||||
exit ${TEST_EXIT_CODE:-0}
|
||||
|
||||
# ==============================================================================
|
||||
# 8-GPU MiniMax-M2.7 (MI30x only)
|
||||
# ==============================================================================
|
||||
@@ -2415,9 +2383,7 @@ jobs:
|
||||
- nightly-8-gpu-mi35x-glm52-fp8-rocm720
|
||||
# 8-GPU GLM-5-MXFP4 (MI35x only)
|
||||
- nightly-8-gpu-mi35x-glm5-mxfp4-rocm720
|
||||
# 4-GPU MiniMax-M2.5 (MI35x)
|
||||
- nightly-4-gpu-mi35x-minimax-m25-rocm720
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x)
|
||||
# 4-GPU MiniMax-M3 MXFP8 (MI35x accuracy + performance)
|
||||
- nightly-4-gpu-mi35x-minimax-m3-rocm720
|
||||
# 8-GPU MiniMax-M2.7 (MI30x only)
|
||||
- nightly-8-gpu-minimax-m27-rocm720
|
||||
|
||||
Reference in New Issue
Block a user