[CI/UT][Refactor] move e2e spec decode and deepseek acc test to per pr (#1136)

### What this PR does / why we need it?
1. run deepseek acc ut per pr --- multicard CI time increased by 9 min
2. run spec decode e2e test on v1 per pr --- singlecard CI time
increased by 3 min (partly is disabled due to not work now)
~~3. align the output of whether dbo is enabled or not~~
    The generated results with and without dbo cannot be aligned.

https://github.com/vllm-project/vllm-ascend/actions/runs/15822900528/job/44600029405?pr=1136
4. skip V0 mtp test due to failure in
https://github.com/vllm-project/vllm-ascend/actions/runs/16012172833/job/45171988816
5. fix some version conflicts
### How was this patch tested?
CI passed with new added test.

---------

Signed-off-by: MengqingCao <cmq0113@163.com>

Author: MengqingCao

.github/workflows/vllm_ascend_test.yaml

@@ -267,7 +267,13 @@ jobs:
           --ignore=tests/e2e/singlecard/test_ilama_lora.py \
           --ignore=tests/e2e/singlecard/test_guided_decoding.py \
           --ignore=tests/e2e/singlecard/test_camem.py \
-          --ignore=tests/e2e/singlecard/test_embedding.py
+          --ignore=tests/e2e/singlecard/test_embedding.py \
+          --ignore=tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py \
+          --ignore=tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py
+          # ------------------------------------ v1 spec decode test ------------------------------------ #
+          VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py
+          # TODO: revert me when test_v1_spec_decode.py::test_ngram_correctness is fixed
+          VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/singlecard/spec_decode_v1/test_v1_spec_decode.py
 
       - name: Run e2e test on V0 engine
         if: ${{ github.event_name == 'schedule' }}
@@ -287,8 +293,6 @@ jobs:
             --ignore=tests/e2e/singlecard/test_guided_decoding.py \
             --ignore=tests/e2e/singlecard/test_camem.py \
             --ignore=tests/e2e/singlecard/test_prompt_embedding.py \
-            --ignore=tests/e2e/singlecard/core/test_ascend_scheduler.py \
-            --ignore=tests/e2e/singlecard/core/test_ascend_scheduler_e2e.py \
             --ignore=tests/e2e/singlecard/test_embedding.py
 
   e2e-4-cards:
@@ -359,7 +363,6 @@ jobs:
           # To avoid oom, we need to run the test in a single process.
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_multistream_moe
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ
-          pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_topk
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W8A8
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_dbo
@@ -379,7 +382,6 @@ jobs:
           # Fixme: run VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py will raise error.
           # To avoid oom, we need to run the test in a single process.
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_QwQ
-          pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_topk
           pytest -sv tests/e2e/multicard/test_offline_inference_distributed.py::test_models_distributed_DeepSeek_W8A8
           pytest -sv tests/e2e/multicard/test_data_parallel.py

.github/workflows/vllm_ascend_test_long_term.yaml

@@ -97,12 +97,9 @@ jobs:
         run: |
           if [[ "${{ matrix.os }}" == "linux-arm64-npu-1" ]]; then
             # v0 spec decode test
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode_v0/e2e/test_mtp_correctness.py  # it needs a clean process
+            # TODO: Revert me when test_mtp_correctness is fixed
+            # VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode_v0/e2e/test_mtp_correctness.py  # it needs a clean process
             pytest -sv tests/e2e/long_term/spec_decode_v0 --ignore=tests/e2e/long_term/spec_decode_v0/e2e/test_mtp_correctness.py
-            # v1 spec decode test
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode_v1/test_v1_mtp_correctness.py
-            # TODO: revert me when test_v1_spec_decode.py::test_ngram_correctness is fixed
-            VLLM_USE_MODELSCOPE=True pytest -sv tests/e2e/long_term/spec_decode_v1/test_v1_spec_decode.py
             # accuracy test single card
             pytest -sv tests/e2e/long_term/test_accuracy.py
           else

tests/e2e/long_term/test_deepseek_v2_lite_tp2_accuracy.py renamed to tests/e2e/multicard/test_deepseek_v2_lite_tp2_accuracy.py

@@ -73,21 +73,6 @@ def test_models_distributed_DeepSeek_multistream_moe():
         vllm_model.generate_greedy(example_prompts, max_tokens)
 
 
-def test_models_distributed_DeepSeek():
-    example_prompts = [
-        "Hello, my name is",
-    ]
-    dtype = "half"
-    max_tokens = 5
-    with VllmRunner(
-            "deepseek-ai/DeepSeek-V2-Lite",
-            dtype=dtype,
-            tensor_parallel_size=4,
-            distributed_executor_backend="mp",
-    ) as vllm_model:
-        vllm_model.generate_greedy(example_prompts, max_tokens)
-
-
 @patch.dict(os.environ, {"VLLM_ASCEND_ENABLE_TOPK_OPTIMIZE": "1"})
 def test_models_distributed_topk() -> None:
     example_prompts = [

tests/e2e/multicard/test_offline_inference_distributed.py

@@ -50,6 +50,8 @@ def model_name():
     return "wemaster/deepseek_mtp_main_random_bf16"
 
 
+@pytest.mark.skipif(
+    True, reason="TODO: Enable me after test_mtp_correctness is fixed")
 def test_mtp_correctness(
     monkeypatch: pytest.MonkeyPatch,
     test_prompts: list[list[dict[str, Any]]],

tests/e2e/long_term/spec_decode_v1/test_v1_mtp_correctness.py renamed to tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py

@@ -314,7 +314,7 @@ def add_request(
         self.block_table.add_row(request.block_ids, req_index)
 
         if sampling_params := request.sampling_params:
-            if (self.is_spec_decode
+            if ((not vllm_version_is("0.9.1")) and self.is_spec_decode
                     and is_spec_decode_unsupported(sampling_params)):
                 self.spec_decode_unsupported_reqs.add(req_id)
             if sampling_params.sampling_type == SamplingType.GREEDY: