diff --git a/.github/workflows/nightly-test-amd-rocm720.yml b/.github/workflows/nightly-test-amd-rocm720.yml index d6dfe0ebe..16f0d3088 100644 --- a/.github/workflows/nightly-test-amd-rocm720.yml +++ b/.github/workflows/nightly-test-amd-rocm720.yml @@ -70,6 +70,7 @@ on: - nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720 - nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720 - nightly-8-gpu-mi35x-deepseek-v4-pro-mtp-rocm720 + - nightly-8-gpu-mi35x-deepseek-v4-pro-dspark-rocm720 # 8-GPU Kimi-K2.6 (MI30x + MI35x) - nightly-8-gpu-kimi-k26-rocm720 - nightly-8-gpu-mi35x-kimi-k26-rocm720 @@ -1438,6 +1439,44 @@ jobs: python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-pro-mtp --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} echo "$(> $GITHUB_STEP_SUMMARY || true + nightly-8-gpu-mi35x-deepseek-v4-pro-dspark-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-8-gpu-mi35x-deepseek-v4-pro-dspark-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 (ROCm 7.2) + run: | + touch github_summary.md + bash scripts/ci/amd/amd_ci_start_container.sh --rocm-version rocm720 + env: + GITHUB_WORKSPACE: ${{ github.workspace }} + ENABLE_CACHE_HOST: "1" + + - name: Install dependencies + run: | + # --skip-test-time-deps: GSM8K doesn't need lmms-eval / human-eval. + bash scripts/ci/amd/amd_ci_install_dependency.sh --skip-test-time-deps + bash scripts/ci/amd/amd_ci_exec.sh pip install tabulate + + - name: Accuracy + DSpark Accept Length Test MI35x ROCm 7.2 (8-GPU DeepSeek-V4-Pro-DSpark FP4, unified_kv_triton) + timeout-minutes: 300 + run: | + > github_summary.md # Clear summary file + echo "## SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton" >> github_summary.md + bash scripts/ci/amd/amd_ci_exec.sh -w /sglang-checkout/test \ + -e SGLANG_HACK_FLASHMLA_BACKEND=unified_kv_triton \ + -e SGLANG_MOE_COPY_WEIGHT_VIEWS_BEFORE_H2D=1 \ + -e GITHUB_STEP_SUMMARY="/sglang-checkout/github_summary.md" \ + python3 run_suite.py --hw amd --suite nightly-amd-8-gpu-mi35x-deepseek-v4-pro-dspark --nightly --timeout-per-file 7200 ${{ (github.event_name == 'schedule' || inputs.continue_on_error) && '--continue-on-error' || '' }} + echo "$(> $GITHUB_STEP_SUMMARY || true + # ============================================================================== # 8-GPU Kimi-K2.6 (MI30x + MI35x) # ============================================================================== @@ -2077,6 +2116,7 @@ jobs: - nightly-8-gpu-mi35x-deepseek-v4-flash-rocm720 - nightly-8-gpu-mi35x-deepseek-v4-pro-rocm720 - nightly-8-gpu-mi35x-deepseek-v4-pro-mtp-rocm720 + - nightly-8-gpu-mi35x-deepseek-v4-pro-dspark-rocm720 # 8-GPU Kimi-K2.6 (MI30x + MI35x) - nightly-8-gpu-kimi-k26-rocm720 - nightly-8-gpu-mi35x-kimi-k26-rocm720