[V1] Support ngram spec decode

Signed-off-by: ponix-j <657511300@qq.com>

Author: ponix-j

.github/workflows/vllm_ascend_test.yaml

@@ -51,11 +51,11 @@ jobs:
         vllm_verison: [main, v0.8.5.post1]
     concurrency:
       group: >
-        ${{ 
-        matrix.os == 'linux-arm64-npu-4' 
-          && github.event.pull_request.number 
-          && format('pr-{0}-limit-npu-4', github.event.pull_request.number) 
-        || format('job-{0}-{1}-{2}', matrix.os, matrix.vllm_verison, github.event.pull_request.number) 
+        ${{
+        matrix.os == 'linux-arm64-npu-4'
+          && github.event.pull_request.number
+          && format('pr-{0}-limit-npu-4', github.event.pull_request.number)
+        || format('job-{0}-{1}-{2}', matrix.os, matrix.vllm_verison, github.event.pull_request.number)
         }}
       cancel-in-progress: false
     name: vLLM Ascend test
@@ -157,8 +157,9 @@ jobs:
         if: steps.filter_spec_decode.outputs.speculative_tests_changed == 'true' || github.event_name == 'schedule'
         run: |
           if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
+            VLLM_USE_MODELSCOPE=true pytest -sv tests/singlecard/spec_decode/e2e/test_v1_spec_decode.py
             pytest -sv tests/singlecard/spec_decode/e2e/test_mtp_correctness.py  # it needs a clean process
-            pytest -sv tests/singlecard/spec_decode --ignore=tests/singlecard/spec_decode/e2e/test_mtp_correctness.py
+            pytest -sv tests/singlecard/spec_decode --ignore=tests/singlecard/spec_decode/e2e/test_mtp_correctness.py --ignore=tests/singlecard/spec_decode/e2e/test_v1_spec_decode.py
           fi
 
       - name: Run vllm-project/vllm test for V0 Engine

requirements-dev.txt

@@ -9,3 +9,4 @@ ray
 types-jsonschema
 xgrammar
 zmq
+numba