[CI] add nvfp4 ci test for b200;

.github/workflows/pr-test-multimodal-gen.yml

@@ -9,6 +9,9 @@ on:
       sgl_kernel:
         required: true
         type: string
+      b200_runner:
+        required: true
+        type: string
       continue_on_error:
         required: false
         type: string
@@ -156,6 +159,53 @@ jobs:
         with:
           artifact-suffix: ${{ matrix.part }}
 
+  multimodal-gen-test-b200:
+    if: |
+      (inputs.target_stage == 'multimodal-gen-test-b200') ||
+      (
+        !inputs.target_stage &&
+        ((github.event_name == 'schedule' || inputs.test_parallel_dispatch == 'true') || (inputs.caller_needs_failure != 'true' && !cancelled())) &&
+        inputs.multimodal_gen == 'true'
+      )
+    runs-on: ${{ inputs.b200_runner }}
+    timeout-minutes: 240
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v4
+        with:
+          ref: ${{ inputs.pr_head_sha || inputs.git_ref || github.sha }}
+
+      - uses: ./.github/actions/check-stage-health
+
+      - uses: ./.github/actions/check-maintenance
+
+      - name: Download artifacts
+        if: inputs.sgl_kernel == 'true'
+        uses: actions/download-artifact@v4
+        with:
+          path: sgl-kernel/dist/
+          merge-multiple: true
+          pattern: wheel-python3.10-cuda12.9
+
+      - name: Install dependencies
+        timeout-minutes: 20
+        run: |
+          CUSTOM_BUILD_SGL_KERNEL=${{inputs.sgl_kernel}} bash scripts/ci/cuda/ci_install_dependency.sh diffusion
+
+      - name: Run diffusion server tests
+        timeout-minutes: 240
+        env:
+          RUNAI_STREAMER_MEMORY_LIMIT: 0
+          CONTINUE_ON_ERROR_FLAG: ${{ inputs.continue_on_error == 'true' && '--continue-on-error' || '' }}
+        run: |
+          cd python
+          python3 sglang/multimodal_gen/test/run_suite.py \
+            --suite 1-gpu-b200 \
+            $CONTINUE_ON_ERROR_FLAG
+
+      - uses: ./.github/actions/upload-cuda-coredumps
+        if: always()
+
   multimodal-gen-unit-test:
     if: |
       (inputs.target_stage == 'multimodal-gen-unit-test') ||

.github/workflows/pr-test.yml

@@ -876,6 +876,7 @@ jobs:
       (
         inputs.target_stage == 'multimodal-gen-test-1-gpu' ||
         inputs.target_stage == 'multimodal-gen-test-2-gpu' ||
+        inputs.target_stage == 'multimodal-gen-test-b200' ||
         inputs.target_stage == 'multimodal-gen-unit-test' ||
         (
           !inputs.target_stage &&
@@ -887,6 +888,7 @@ jobs:
     with:
       multimodal_gen: ${{ needs.check-changes.outputs.multimodal_gen }}
       sgl_kernel: ${{ needs.check-changes.outputs.sgl_kernel }}
+      b200_runner: ${{ needs.check-changes.outputs.b200_runner }}
       continue_on_error: ${{ needs.check-changes.outputs.continue_on_error }}
       pr_head_sha: ${{ inputs.pr_head_sha || '' }}
       git_ref: ${{ inputs.git_ref || '' }}

python/sglang/multimodal_gen/test/run_suite.py

@@ -50,6 +50,9 @@
         "test_server_2_gpu_b.py",
         # add new 2-gpu test files here
     ],
+    "1-gpu-b200": [
+        "test_server_c.py",
+    ],
 }
 
 suites_ascend = {
@@ -78,7 +81,7 @@ def parse_args():
         type=str,
         required=True,
         choices=list(SUITES.keys()),
-        help="The test suite to run (e.g., 1-gpu, 2-gpu)",
+        help="The test suite to run (valid names are defined in SUITES)",
     )
     parser.add_argument(
         "--partition-id",

python/sglang/multimodal_gen/test/scripts/gen_diffusion_ci_outputs.py

@@ -8,6 +8,7 @@
 Usage:
     python gen_diffusion_ci_outputs.py --suite 1-gpu --partition-id 0 --total-partitions 2 --out-dir ./output
     python gen_diffusion_ci_outputs.py --suite 1-gpu --case-ids qwen_image_t2i flux_image_t2i --out-dir ./output
+    python gen_diffusion_ci_outputs.py --suite 1-gpu-b200 --out-dir ./output
 """
 
 import argparse
@@ -27,9 +28,9 @@ def main():
     parser.add_argument(
         "--suite",
         type=str,
-        choices=["1-gpu", "2-gpu"],
+        choices=["1-gpu", "2-gpu", "1-gpu-b200"],
         required=True,
-        help="Test suite to run (1-gpu or 2-gpu)",
+        help="Test suite to run (1-gpu, 2-gpu, or 1-gpu-b200)",
     )
     parser.add_argument(
         "--partition-id",

python/sglang/multimodal_gen/test/server/test_server_c.py

@@ -0,0 +1,28 @@
+"""
+Config-driven diffusion performance test with pytest parametrization.
+"""
+
+from __future__ import annotations
+
+import pytest
+
+from sglang.multimodal_gen.runtime.utils.logging_utils import init_logger
+from sglang.multimodal_gen.test.server.test_server_common import (  # noqa: F401
+    DiffusionServerBase,
+    diffusion_server,
+)
+from sglang.multimodal_gen.test.server.testcase_configs import (
+    ONE_GPU_CASES_C,
+    DiffusionTestCase,
+)
+
+logger = init_logger(__name__)
+
+
+class TestDiffusionServerOneGpuB200(DiffusionServerBase):
+    """B200-targeted smoke tests for 1-GPU diffusion cases."""
+
+    @pytest.fixture(params=ONE_GPU_CASES_C, ids=lambda c: c.id)
+    def case(self, request) -> DiffusionTestCase:
+        """Provide a DiffusionTestCase for each 1-GPU B200 test."""
+        return request.param

python/sglang/multimodal_gen/test/server/testcase_configs.py

@@ -798,7 +798,7 @@ def from_req_perf_record(
         )
     )
 
-# TODO: enable on 4090/5090/b200
+# TODO: enable on 4090/5090
 ONE_GPU_CASES_C = [
     DiffusionTestCase(
         "flux_2_nvfp4_t2i",
@@ -807,6 +807,7 @@ def from_req_perf_record(
             modality="image",
         ),
         T2I_sampling_params,
+        run_perf_check=False,
     )
 ]
 

scripts/ci/utils/slash_command_handler.py

@@ -271,6 +271,7 @@ def handle_rerun_stage(
         "stage-c-test-deepep-8-gpu-h200",
         "multimodal-gen-test-1-gpu",
         "multimodal-gen-test-2-gpu",
+        "multimodal-gen-test-b200",
     ]
 
     # Valid AMD stage names that support target_stage