[NPU][CI] Add CANN 9.1.0 and Ascend a5 nightly suites (#38833)

.github/workflows/_npu-single-node-test-stage.yml

@@ -345,6 +345,7 @@ jobs:
           source /usr/local/Ascend/nnal/atb/set_env.sh || true
           source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/customize/bin/set_env.bash || true
           source /usr/local/Ascend/ascend-toolkit/latest/opp/vendors/custom_transformer/bin/set_env.bash || true
+          source /usr/local/Ascend/ascend-toolkit/latest/set_env.sh || true
           set -u
 
           # Nightly log path mirrors the output layout; PR keeps the original {date}/{testcase}.

.github/workflows/nightly-test-npu.yml

@@ -29,6 +29,16 @@ on:
         required: false
         type: string
         default: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:main-cann9.0.0-a3'
+      image_a3_cann910:
+        description: 'The a3 running docker image of the test task with cann 9.1.0.'
+        required: false
+        type: string
+        default: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:cann9.1.0-a3-nightly'
+      image_a5:
+        description: 'The a5 running docker image of the test task.'
+        required: false
+        type: string
+        default: 'swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:cann9.1.0-950-nightly'
       skip_install_flag:
         description: 'Indicates whether to skip the installation of sglang, defaulting to false.'
         required: false
@@ -56,6 +66,16 @@ on:
         required: false
         type: string
         default: ''
+      image_a3_cann910:
+        description: 'The a3 running docker image of the test task with cann 9.1.0.'
+        required: false
+        type: string
+        default: ''
+      image_a5:
+        description: 'The a5 running docker image of the test task.'
+        required: false
+        type: string
+        default: ''
       skip_install_flag:
         description: 'Indicates whether to skip the installation of sglang, defaulting to false.'
         required: false
@@ -72,6 +92,8 @@ jobs:
       job_filter: ${{ steps.set-vars.outputs.job_filter }}
       image_a2: ${{ steps.set-vars.outputs.image_a2 }}
       image_a3: ${{ steps.set-vars.outputs.image_a3 }}
+      image_a3_cann910: ${{ steps.set-vars.outputs.image_a3_cann910 }}
+      image_a5: ${{ steps.set-vars.outputs.image_a5 }}
       skip_install_flag: ${{ steps.set-vars.outputs.skip_install_flag }}
       run_start_metadata: ${{ steps.set-vars.outputs.run_start_metadata }}
     steps:
@@ -103,6 +125,16 @@ jobs:
           else
             echo "image_a3=${{ inputs.image_a3 }}" >> $GITHUB_OUTPUT
           fi
+          if [ -z "${{ inputs.image_a3_cann910 }}" ]; then
+            echo "image_a3_cann910=swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:cann9.1.0-a3-nightly" >> $GITHUB_OUTPUT
+          else
+            echo "image_a3_cann910=${{ inputs.image_a3_cann910 }}" >> $GITHUB_OUTPUT
+          fi
+          if [ -z "${{ inputs.image_a5 }}" ]; then
+            echo "image_a5=swr.cn-southwest-2.myhuaweicloud.com/base_image/dockerhub/lmsysorg/sglang:cann9.1.0-950-nightly" >> $GITHUB_OUTPUT
+          else
+            echo "image_a5=${{ inputs.image_a5 }}" >> $GITHUB_OUTPUT
+          fi
           if [ -z "${{ inputs.skip_install_flag }}" ]; then
             echo "skip_install_flag=false" >> $GITHUB_OUTPUT
           else
@@ -282,6 +314,56 @@ jobs:
       device_type_for_deps: 'a3'
       run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
 
+  nightly-acc-4-npu-a5:
+    name: nightly-acc-4-npu-a5
+    if: ${{ !cancelled() }}
+    needs: [set-image-config]
+    uses: ./.github/workflows/_npu-single-node-test-stage.yml
+    with:
+      runner: linux-amd64-a5-4
+      test_type: 'accuracy'
+      test_suite: nightly-acc-4-npu-a5
+      is_nightly_pipeline_job: true
+      skip_pr_test_health_check: 'true'
+      image: ${{ needs.set-image-config.outputs.image_a5 }}
+      install_sglang_deps: false
+      device_type_for_deps: 'a5'
+      run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
+
+  # Suites suffixed with `-cann910` run against the CANN 9.1.0 image.
+  nightly-perf-16-npu-a3-cann910:
+    name: nightly-perf-16-npu-a3-cann910
+    if: ${{ !cancelled() }}
+    needs: [set-image-config]
+    uses: ./.github/workflows/_npu-single-node-test-stage.yml
+    with:
+      runner: linux-aarch64-a3-800t-16
+      test_type: 'perf'
+      test_suite: nightly-perf-16-npu-a3-cann910
+      is_nightly_pipeline_job: true
+      skip_pr_test_health_check: 'true'
+      image: ${{ needs.set-image-config.outputs.image_a3_cann910 }}
+      install_sglang_deps: false
+      device_type_for_deps: 'a3'
+      run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
+      test_timeout_minutes: '180'
+
+  nightly-acc-16-npu-a3-cann910:
+    name: nightly-acc-16-npu-a3-cann910
+    if: ${{ !cancelled() }}
+    needs: [set-image-config]
+    uses: ./.github/workflows/_npu-single-node-test-stage.yml
+    with:
+      runner: linux-aarch64-a3-16-
+      test_type: 'accuracy'
+      test_suite: nightly-acc-16-npu-a3-cann910
+      is_nightly_pipeline_job: true
+      skip_pr_test_health_check: 'true'
+      image: ${{ needs.set-image-config.outputs.image_a3_cann910 }}
+      install_sglang_deps: false
+      device_type_for_deps: 'a3'
+      run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
+
   nightly-poc-multi-node-tests:
     name: multi-node-poc
     if: ${{ !cancelled() }}
@@ -362,6 +444,72 @@ jobs:
       transformers_version: ''
       run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
 
+  nightly-poc-multi-node-tests-cann910:
+    name: multi-node-poc-cann910
+    if: ${{ !cancelled() }}
+    # Multi-node jobs run strictly one at a time to avoid contending for the
+    # same A3 multi-node resources: 9.0.0 multi -> 9.0.0 mix -> cann910 multi -> cann910 mix.
+    needs: [set-image-config, nightly-perf-16-npu-a3-cann910, nightly-acc-16-npu-a3-cann910, nightly-poc-multi-node-tests, nightly-poc-multi-node-mix-tests]
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        test_config:
+          # deepseek_v4_flash performance tests
+          - name: deepseek_v4_flash_w8a8_1p1d_16p_in8k_out1k_50ms
+            prefill_size: 1
+            decode_size: 1
+            router_size: 1
+            test_case: test/registered/npu/performance/deepseek_v4_flash/test_npu_deepseek_v4_flash_w8a8_1p1d_16p_in8k_out1k_50ms.py
+            test_type: 'perf'
+            prefill_decode_deployment: 'separation'
+    uses: ./.github/workflows/nightly-test-npu-e2e-multi-node.yml
+    with:
+      runner: linux-amd64-cpu-4
+      test_type: ${{ matrix.test_config.test_type || 'perf' }}
+      test_config_name: ${{ matrix.test_config.name }}
+      prefill_size: ${{ matrix.test_config.prefill_size }}
+      decode_size: ${{ matrix.test_config.decode_size }}
+      router_size: ${{ matrix.test_config.router_size }}
+      test_case: ${{ matrix.test_config.test_case }}
+      image: ${{ needs.set-image-config.outputs.image_a3_cann910 }}
+      install_sglang_from_source: false
+      prefill_decode_deployment: ${{ matrix.test_config.prefill_decode_deployment }}
+      transformers_version: ''
+      run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
+
+  nightly-poc-multi-node-mix-tests-cann910:
+    name: multi-node-mix-poc-cann910
+    if: ${{ !cancelled() }}
+    needs: [set-image-config, nightly-poc-multi-node-tests-cann910]
+    strategy:
+      fail-fast: false
+      max-parallel: 1
+      matrix:
+        test_config:
+          # glm_5_2 accuracy tests
+          - name: glm_5_2_w4a8_16p_gpqa
+            node_size: 2
+            test_case: test/registered/npu/accuracy/glm5_2/test_npu_glm_5_2_w4a8_16p_gpqa.py
+            test_type: 'accuracy'
+          # kimi_k3 accuracy tests
+          - name: kimi_k3_w4a8_32p_gpqa
+            node_size: 4
+            test_case: test/registered/npu/accuracy/kimi_k3/test_npu_kimi_k3_w4a8_32p_gpqa.py
+            test_type: 'accuracy'
+    uses: ./.github/workflows/nightly-test-npu-e2e-multi-node.yml
+    with:
+      runner: linux-amd64-cpu-4
+      test_type: ${{ matrix.test_config.test_type || 'perf' }}
+      test_config_name: ${{ matrix.test_config.name }}
+      node_size: ${{ matrix.test_config.node_size }}
+      test_case: ${{ matrix.test_config.test_case }}
+      image: ${{ needs.set-image-config.outputs.image_a3_cann910 }}
+      install_sglang_from_source: false
+      prefill_decode_deployment: 'mix'
+      transformers_version: ''
+      run_start_metadata: ${{ needs.set-image-config.outputs.run_start_metadata }}
+
   check-all-jobs:
     if: ${{ !cancelled() }}
     needs:
@@ -375,14 +523,21 @@ jobs:
       - nightly-perf-16-npu-a3
       - nightly-acc-2-npu-a3
       - nightly-acc-16-npu-a3
+      - nightly-perf-16-npu-a3-cann910
+      - nightly-acc-16-npu-a3-cann910
+      - nightly-acc-4-npu-a5
       - nightly-poc-multi-node-tests
       - nightly-poc-multi-node-mix-tests
+      - nightly-poc-multi-node-tests-cann910
+      - nightly-poc-multi-node-mix-tests-cann910
     runs-on: ubuntu-latest
     steps:
       - name: Generate results table
         run: |
           multi_result="${{ needs.nightly-poc-multi-node-tests.result }}"
           mix_result="${{ needs.nightly-poc-multi-node-mix-tests.result }}"
+          multi_cann910_result="${{ needs.nightly-poc-multi-node-tests-cann910.result }}"
+          mix_cann910_result="${{ needs.nightly-poc-multi-node-mix-tests-cann910.result }}"
 
           # single-node suites are reported per suite; overall status is success if none failed
           single_result="success"
@@ -396,7 +551,10 @@ jobs:
             "${{ needs.nightly-perf-4-npu-a3.result }}" \
             "${{ needs.nightly-perf-16-npu-a3.result }}" \
             "${{ needs.nightly-acc-2-npu-a3.result }}" \
-            "${{ needs.nightly-acc-16-npu-a3.result }}"; do
+            "${{ needs.nightly-acc-16-npu-a3.result }}" \
+            "${{ needs.nightly-perf-16-npu-a3-cann910.result }}" \
+            "${{ needs.nightly-acc-16-npu-a3-cann910.result }}" \
+            "${{ needs.nightly-acc-4-npu-a5.result }}"; do
             if [ "${r}" != "success" ] && [ "${r}" != "skipped" ]; then
               single_result="failure"
             fi
@@ -426,17 +584,22 @@ jobs:
             "nightly-perf-4-npu-a3:${{ needs.nightly-perf-4-npu-a3.result }}" \
             "nightly-perf-16-npu-a3:${{ needs.nightly-perf-16-npu-a3.result }}" \
             "nightly-acc-2-npu-a3:${{ needs.nightly-acc-2-npu-a3.result }}" \
-            "nightly-acc-16-npu-a3:${{ needs.nightly-acc-16-npu-a3.result }}"; do
+            "nightly-acc-16-npu-a3:${{ needs.nightly-acc-16-npu-a3.result }}" \
+            "nightly-perf-16-npu-a3-cann910:${{ needs.nightly-perf-16-npu-a3-cann910.result }}" \
+            "nightly-acc-16-npu-a3-cann910:${{ needs.nightly-acc-16-npu-a3-cann910.result }}" \
+            "nightly-acc-4-npu-a5:${{ needs.nightly-acc-4-npu-a5.result }}"; do
             suite="${entry%%:*}"
             r="${entry##*:}"
             echo "| ${suite} | $(group_icon ${r}) ${r} |" >> $GITHUB_STEP_SUMMARY
           done
           echo "| multi-node-poc | $(group_icon ${multi_result}) ${multi_result} |" >> $GITHUB_STEP_SUMMARY
           echo "| multi-node-mix-poc | $(group_icon ${mix_result}) ${mix_result} |" >> $GITHUB_STEP_SUMMARY
+          echo "| multi-node-poc-cann910 | $(group_icon ${multi_cann910_result}) ${multi_cann910_result} |" >> $GITHUB_STEP_SUMMARY
+          echo "| multi-node-mix-poc-cann910 | $(group_icon ${mix_cann910_result}) ${mix_cann910_result} |" >> $GITHUB_STEP_SUMMARY
           echo "" >> $GITHUB_STEP_SUMMARY
 
           FAIL=0
-          for r in "${single_result}" "${multi_result}" "${mix_result}"; do
+          for r in "${single_result}" "${multi_result}" "${mix_result}" "${multi_cann910_result}" "${mix_cann910_result}"; do
             if [ "${r}" != "success" ] && [ "${r}" != "skipped" ]; then FAIL=1; fi
           done
           exit $FAIL

python/sglang/test/ascend/e2e/test_npu_performance_utils.py

@@ -183,6 +183,9 @@ QWEN3_5_397B_W8A8_MODEL_PATH = (
 DEEPSEEK_V4_FLASH_0731_W8A8_MODEL_PATH = (
     "/root/.cache/modelscope/hub/models/Eco-Tech/DeepSeek-V4-Flash-0731-w8a8"
 )
+DEEPSEEK_V4_FLASH_DEFAULT_MODEL_PATH = (
+    "/root/.cache/modelscope/hub/models/deepseek-ai/DeepSeek-V4-Flash"
+)
 QWEN3_5_397B_W4A8_MODEL_PATH = (
     "/root/.cache/modelscope/hub/models/Eco-Tech/Qwen3.5-397B-A17B-w4a8-mtp"
 )

test/registered/npu/accuracy/deepseek_v4_flash/test_npu_deepseek_v4_flash_fp8_4p_gpqa_a5.py

@@ -0,0 +1,143 @@
+import unittest
+
+from sglang.test.ascend.e2e.test_npu_accuracy_utils import (
+    BENCHMARK_TOOL_DEFAULT,
+    TestNpuAccuracyTestCaseBase,
+)
+from sglang.test.ascend.e2e.test_npu_performance_utils import (
+    DEEPSEEK_V4_FLASH_DEFAULT_MODEL_PATH,
+)
+from sglang.test.ci.ci_register import register_npu_ci
+
+register_npu_ci(
+    est_time=7200,
+    suite="nightly-acc-4-npu-a5",
+    nightly=True,
+)
+
+DEEPSEEK_V4_FLASH_FP8_4P_ENVS = {
+    "PYTORCH_NPU_ALLOC_CONF": "expandable_segments:True",
+    "STREAMS_PER_DEVICE": "32",
+    "SGLANG_SET_CPU_AFFINITY": "1",
+    "TASK_QUEUE_ENABLE": "1",
+    "INF_NAN_MODE_FORCE_DISABLE": "1",
+    "SGLANG_DEFAULT_THINKING": "1",
+    "SGLANG_DSV4_REASONING_EFFORT": "max",
+    # HCCL deepep
+    "HCCL_BUFFSIZE": "1024",
+    "HCCL_SOCKET_IFNAME": "lo",
+    "GLOO_SOCKET_IFNAME": "lo",
+    # DeepEP long-sequence normal combine (ROUND * TOKENS >= chunked_prefill_size / tp * dp)
+    "DEEPEP_NORMAL_COMBINE_ENABLE_LONG_SEQ": "1",
+    "DEEPEP_NORMAL_LONG_SEQ_ROUND": "16",
+    "DEEPEP_NORMAL_LONG_SEQ_PER_ROUND_TOKENS": "2048",
+    # "DEEPEP_HYBRID_DEPLOYMENT": "1",
+    # dsv4
+    "IS_DEEPSEEK_V4": "1",
+    "USE_FUSED_HC_PRE_ASCENDC": "1",
+    "SGLANG_DSV4_NPU_FUSED_COMPRESSOR": "1",
+    "SGLANG_DSV4_NPU_FUSED_COMPRESSOR_PREFILL": "1",
+    # skip gpu branch
+    "SGLANG_OPT_USE_OVERLAP_STORE_CACHE": "False",
+    "FORCE_DRAFT_MODEL_NON_QUANT": "1",
+    "SGLANG_DSV4_FP4_EXPERTS": "True",
+    "SGLANG_OPT_FUSE_WQA_WKV": "0",
+    "SGLANG_OPT_BF16_FP32_GEMM_ALGO": "torch",
+    "SGLANG_OPT_USE_FUSED_HASH_TOPK": "False",
+    "SGLANG_OPT_USE_TILELANG_MHC_PRE": "False",
+    "SGLANG_OPT_DEEPGEMM_HC_PRENORM": "False",
+    "SGLANG_OPT_USE_TILELANG_MHC_POST": "False",
+    "SGLANG_OPT_FP8_WO_A_GEMM": "False",
+    # performance
+    "SGLANG_ENABLE_OVERLAP_PLAN_STREAM": "1",
+    "SGLANG_NPU_USE_MULTI_STREAM": "1",
+    "USE_NPU_MOE_GATING_TOP_K": "1",
+    "SGLANG_PROFILE_WITH_STACK": "False",
+}
+
+DEEPSEEK_V4_FLASH_FP8_4P_OTHER_ARGS = [
+    "--page-size",
+    128,
+    "--tp-size",
+    4,
+    "--trust-remote-code",
+    "--attention-backend",
+    "dsv4",
+    "--device",
+    "npu",
+    "--watchdog-timeout",
+    9000,
+    "--mem-fraction-static",
+    0.72,
+    "--max-running-requests",
+    64,
+    "--chunked-prefill-size",
+    131072,
+    "--max-prefill-tokens",
+    131072,
+    "--cuda-graph-bs-decode",
+    1,
+    2,
+    4,
+    8,
+    10,
+    16,
+    "--kv-cache-dtype",
+    "auto",
+    "--enable-dp-lm-head",
+    "--disable-radix-cache",
+    "--enable-dp-attention",
+    "--dp-size",
+    4,
+    "--reasoning-parser",
+    "deepseek-v4",
+    "--speculative-algorithm",
+    "EAGLE",
+    "--speculative-num-steps",
+    2,
+    "--speculative-eagle-topk",
+    1,
+    "--speculative-num-draft-tokens",
+    3,
+    "--quantization",
+    "fp8",
+    "--moe-a2a-backend",
+    "deepep",
+    "--deepep-mode",
+    "auto",
+]
+
+DEEPSEEK_V4_FLASH_FP8_4P_GENERATION_CONFIG_MAX = {
+    "max_tokens": 125000,
+    "top_p": 1,
+    "temperature": 1,
+    "n": 1,
+    "extra_body": {
+        "chat_template_kwargs": {"thinking": True, "reasoning_effort": "max"}
+    },
+}
+
+
+class TestNPUDeepSeekV4FlashFP84PGPQA(TestNpuAccuracyTestCaseBase):
+    """Test NPU accuracy for DeepSeek-V4-Flash FP8 4p EAGLE GPQA."""
+
+    benchmark_tool = BENCHMARK_TOOL_DEFAULT
+    model = DEEPSEEK_V4_FLASH_DEFAULT_MODEL_PATH
+    other_args = DEEPSEEK_V4_FLASH_FP8_4P_OTHER_ARGS
+    envs = DEEPSEEK_V4_FLASH_FP8_4P_ENVS
+    accuracy = 0.874
+    datasets = ["gpqa_diamond"]
+    few_shot_num = 0
+    generation_config = DEEPSEEK_V4_FLASH_FP8_4P_GENERATION_CONFIG_MAX
+    eval_batch_size = 128
+    stream = True
+    timeout = 6000
+    seed = 1
+
+    def test_npu_deepseek_v4_flash_fp8_4p_gpqa(self):
+        """Run NPU accuracy test for DeepSeek-V4-Flash FP8 4p EAGLE GPQA."""
+        self.run_accuracy()
+
+
+if __name__ == "__main__":
+    unittest.main()

test/registered/npu/accuracy/deepseek_v4_flash/test_npu_deepseek_v4_flash_w8a8_8p_gpqa.py

@@ -15,6 +15,12 @@ register_npu_ci(
     nightly=True,
 )
 
+register_npu_ci(
+    est_time=7200,
+    suite="nightly-acc-16-npu-a3-cann910",
+    nightly=True,
+)
+
 DEEPSEEK_V4_FLASH_W8A8_DSPARK_8P_ENVS = {
     "PYTORCH_NPU_ALLOC_CONF": "expandable_segments:True",
     "STREAMS_PER_DEVICE": "32",

test/registered/npu/performance/deepseek_v4_flash/test_npu_deepseek_v4_flash_w8a8_8p_in32k_out1k_50ms.py

@@ -1,132 +0,0 @@
-import unittest
-
-from sglang.test.ascend.e2e.test_npu_performance_utils import (
-    AISBENCHMARK_DATASET_DEFAULT,
-    BENCHMARK_TOOL_DEFAULT,
-    DEEPSEEK_V4_FLASH_0731_W8A8_MODEL_PATH,
-    TestNpuPerformanceTestCaseBase,
-)
-from sglang.test.ci.ci_register import register_npu_ci
-
-register_npu_ci(est_time=1800, suite="nightly-perf-16-npu-a3", nightly=True)
-
-# Environment variables for DSV4-Flash single-node PD-mix deployment.
-DEEPSEEK_V4_FLASH_W8A8_8P_ENVS = {
-    "PYTORCH_NPU_ALLOC_CONF": "expandable_segments:True",
-    "STREAMS_PER_DEVICE": "32",
-    "INF_NAN_MODE_FORCE_DISABLE": "1",
-    "SGLANG_SET_CPU_AFFINITY": "1",
-    "HCCL_SOCKET_IFNAME": "lo",
-    "GLOO_SOCKET_IFNAME": "lo",
-    "HCCL_OP_EXPANSION_MODE": "AIV",
-    "USE_NPU_MOE_GATING_TOP_K": "1",
-    "SGLANG_NPU_USE_MULTI_STREAM": "1",
-    # deepep
-    "DEEP_NORMAL_MODE_USE_INT8_QUANT": "1",
-    "DEEPEP_HCCL_BUFFSIZE": "2048",
-    "SGLANG_DEEPEP_NUM_MAX_DISPATCH_TOKENS_PER_RANK": "35",
-    "DEEPEP_HYBRID_DEPLOYMENT": "1",
-    # war barrier
-    "SGLANG_ENABLE_WAR_BARRIER": "1",
-    "SGLANG_FORCE_COARSE_WAR_BARRIER": "1",
-    # skip gpu branch
-    "SGLANG_OPT_FP8_WO_A_GEMM": "0",
-    "SGLANG_OPT_USE_OVERLAP_STORE_CACHE": "False",
-    "FORCE_DRAFT_MODEL_NON_QUANT": "1",
-    "SGLANG_DSV4_FP4_EXPERTS": "False",
-    "SGLANG_OPT_FUSE_WQA_WKV": "0",
-    "SGLANG_OPT_BF16_FP32_GEMM_ALGO": "torch",
-    "SGLANG_OPT_USE_FUSED_HASH_TOPK": "False",
-    "SGLANG_OPT_USE_TILELANG_MHC_PRE": "False",
-    "SGLANG_OPT_DEEPGEMM_HC_PRENORM": "False",
-    "SGLANG_OPT_USE_TILELANG_MHC_POST": "False",
-    # mtp
-    "SGLANG_ENABLE_SPEC_V2": "1",
-    "SGLANG_ENABLE_OVERLAP_PLAN_STREAM": "1",
-}
-
-# Server launch arguments for DSV4-Flash W8A8 single-node 8p PD-mix.
-DEEPSEEK_V4_FLASH_W8A8_8P_OTHER_ARGS = [
-    "--page-size",
-    128,
-    "--tp-size",
-    16,
-    "--trust-remote-code",
-    "--device",
-    "npu",
-    "--prefill-max-requests",
-    32,
-    "--max-prefill-tokens",
-    80000,
-    "--attention-backend",
-    "dsv4",
-    "--watchdog-timeout",
-    9000,
-    "--mem-fraction-static",
-    0.68,
-    "--chunked-prefill-size",
-    131072,
-    "--max-running-requests",
-    64,
-    "--dp-size",
-    16,
-    "--enable-dp-attention",
-    "--moe-a2a-backend",
-    "deepep",
-    "--deepep-mode",
-    "auto",
-    "--quantization",
-    "modelslim",
-    "--enable-dp-lm-head",
-    "--kv-cache-dtype",
-    "bfloat16",
-    "--skip-server-warmup",
-    "--cuda-graph-bs-decode",
-    1,
-    2,
-    4,
-    8,
-    # MTP (EAGLE) configuration.
-    "--speculative-algorithm",
-    "EAGLE",
-    "--speculative-num-steps",
-    2,
-    "--speculative-eagle-topk",
-    1,
-    "--speculative-num-draft-tokens",
-    3,
-    "--ep-size",
-    16,
-    "--disable-radix-cache",
-]
-
-
-class TestNPUDeepSeekV4FlashW8A88PIn32kOut1k50ms(TestNpuPerformanceTestCaseBase):
-    """Test NPU performance for DeepSeek-V4-Flash W8A8 8p in32k out1k."""
-
-    benchmark_tool = BENCHMARK_TOOL_DEFAULT
-    dataset_type = AISBENCHMARK_DATASET_DEFAULT
-    model = DEEPSEEK_V4_FLASH_0731_W8A8_MODEL_PATH
-    other_args = DEEPSEEK_V4_FLASH_W8A8_8P_OTHER_ARGS
-    envs = DEEPSEEK_V4_FLASH_W8A8_8P_ENVS
-    dataset_name = "random"
-    dataset_path = "/root/.cache/modelscope/hub/datasets/gsm8k_deepseekv4/cache0_32000/formal_run1_64_32000_cache0.json"
-    input_len = 32000
-    output_len = 1000
-    num_prompts = 64
-    max_concurrency = 64
-    random_range_ratio = 1
-    warmup_requests = 16
-    request_rate = float("inf")
-    seed = 1
-    tpot = 50
-    max_attempts = 3
-    output_token_throughput = 927
-
-    def test_npu_deepseek_v4_flash_w8a8_8p_in32k_out1k_50ms(self):
-        """Run NPU performance test for DeepSeek-V4-Flash W8A8 8p in32k out1k."""
-        self.run_throughput()
-
-
-if __name__ == "__main__":
-    unittest.main()

test/registered/npu/performance/deepseek_v4_flash/test_npu_deepseek_v4_flash_w8a8_8p_in8k_out1k_50ms.py

@@ -9,6 +9,7 @@ from sglang.test.ascend.e2e.test_npu_performance_utils import (
 from sglang.test.ci.ci_register import register_npu_ci
 
 register_npu_ci(est_time=1800, suite="nightly-perf-16-npu-a3", nightly=True)
+register_npu_ci(est_time=1800, suite="nightly-perf-16-npu-a3-cann910", nightly=True)
 
 # Environment variables for DSV4-Flash single-node PD-mix deployment.
 DEEPSEEK_V4_FLASH_W8A8_8P_ENVS = {

test/run_suite.py

@@ -170,9 +170,12 @@ NIGHTLY_SUITES = {
         "nightly-acc-2-npu-a3",
         "nightly-acc-8-npu-a3",
         "nightly-acc-16-npu-a3",
+        "nightly-acc-4-npu-a5",
         "nightly-perf-2-npu-a3",
         "nightly-perf-4-npu-a3",
         "nightly-perf-16-npu-a3",
+        "nightly-perf-16-npu-a3-cann910",
+        "nightly-acc-16-npu-a3-cann910",
         "full-4-npu-a2",
         "full-1-npu-a3",
         "full-2-npu-a3",