Add GB300 base C CI suite (#27427)

.github/workflows/_pr-test-stage.yml

@@ -119,6 +119,8 @@ jobs:
 
       - name: Install dependencies
         timeout-minutes: ${{ fromJson(steps.rc.outputs.install_timeout) }}
+        env:
+          GRACE_BLACKWELL: ${{ steps.rc.outputs.grace_blackwell || '0' }}
         run: |
           CUSTOM_BUILD_SGL_KERNEL=${{ fromJson(inputs.check_changes).sgl_kernel }} bash ${{ steps.rc.outputs.install }}
 

.github/workflows/pr-test.yml

@@ -59,8 +59,6 @@ env:
   SGLANG_CUDA_COREDUMP: "1"
   SGLANG_JIT_DEEPGEMM_FAST_WARMUP: true
   SKIP_PR_TEST_HEALTH_CHECK: ${{ (inputs.skip_pr_test_health_check == true || inputs.test_parallel_dispatch == true || inputs.run_all_tests == true) && 'true' || 'false' }}
-  # TEMP: rebuild deepep against the new torch for torch-211-merge PR only — revert before merging to main.
-  FORCE_REBUILD_DEEPEP: '1'
   # Schedule / main-branch dispatch / workflow_call from main use refs/heads/main; PR events use refs/pull/*/merge
   PR_TEST_BYPASS_MAINTENANCE_ON_MAIN: ${{ github.ref == 'refs/heads/main' && 'true' || 'false' }}
   USE_VENV: false
@@ -515,6 +513,20 @@ jobs:
       timeout_per_file: '1800'
     secrets: inherit
 
+  base-c-test-4-gpu-gb300:
+    needs: [check-changes, call-gate, wait-for-base-b, sgl-kernel-build-wheels]
+    if: ${{ !failure() && !cancelled() }}
+    uses: ./.github/workflows/_pr-test-stage.yml
+    with:
+      self_name: base-c-test-4-gpu-gb300
+      runner_config: 4-gpu-gb300
+      check_changes: ${{ toJson(needs.check-changes.outputs) }}
+      caller_inputs: ${{ toJson(inputs) }}
+      partitions: ${{ needs.check-changes.outputs.partitions }}
+      run_timeout_minutes: '30'
+      timeout_per_file: '1800'
+    secrets: inherit
+
   pr-test-finish:
     needs:
       [
@@ -545,7 +557,7 @@ jobs:
         base-c-test-deepep-4-gpu-b200,
         base-c-test-deepep-8-gpu-h200,
         base-c-test-4-gpu-b200,
-        # base-c-test-4-gpu-gb200,  # Temporarily disabled — no GB200 runner
+        base-c-test-4-gpu-gb300,
       ]
     if: always()
     runs-on: ubuntu-latest

.github/workflows/rerun-test.yml

@@ -31,6 +31,11 @@ on:
         required: false
         type: string
         default: "20"
+      grace_blackwell:
+        description: "Set GRACE_BLACKWELL for the install step (cuda only)"
+        required: false
+        type: string
+        default: "0"
       rdma_devices:
         description: "SGLANG_CI_RDMA_ALL_DEVICES csv (cuda only; empty = unset)"
         required: false
@@ -100,6 +105,8 @@ jobs:
 
       - name: Install dependencies
         timeout-minutes: ${{ fromJson(inputs.install_timeout) }}
+        env:
+          GRACE_BLACKWELL: ${{ inputs.grace_blackwell }}
         run: |
           if [[ "${{ inputs.runs_on }}" == "1-gpu-5090" ]]; then
             source /etc/profile.d/sglang-ci.sh