remove useless patch (#4699)

wangxiyuan · MengqingCao · web-flow · commit 0b65ac6c4b75 · 2025-12-08T11:02:42.000+08:00
patach_config is useless now. Let's remove it - vLLM version: v0.12.0 - vLLM main: vllm-project/vllm@ad32e3e Signed-off-by: wangxiyuan <wangxiyuan1007@gmail.com> Co-authored-by: Mengqing Cao <cmq0113@163.com>
diff --git a/examples/external_online_dp/run_dp_template.sh b/examples/external_online_dp/run_dp_template.sh
@@ -29,4 +29,4 @@ vllm serve model_path \
     --trust-remote-code \
     --gpu-memory-utilization 0.9 \
     --quantization ascend \
-    --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}' \
+    --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
diff --git a/tests/e2e/nightly/features/test_mtpx_deepseek_r1_0528_w8a8.py b/tests/e2e/nightly/features/test_mtpx_deepseek_r1_0528_w8a8.py
@@ -74,10 +74,7 @@ async def test_models(model: str, mode: str) -> None:
         "VLLM_EXECUTE_MODEL_TIMEOUT_SECONDS": "3600000"
     }
     additional_config: dict[str, Any] = {}
-    speculative_config = {
-        "num_speculative_tokens": 2,
-        "method": "deepseek_mtp"
-    }
+    speculative_config = {"num_speculative_tokens": 2, "method": "mtp"}
     compilation_config = {
         "cudagraph_capture_sizes": [56],
         "cudagraph_mode": "FULL_DECODE_ONLY"
diff --git a/tests/e2e/nightly/features/test_prefix_cache_deepseek_r1_0528_w8a8.py b/tests/e2e/nightly/features/test_prefix_cache_deepseek_r1_0528_w8a8.py
@@ -84,10 +84,7 @@ async def test_models(model: str) -> None:
         "chunked_prefill_for_mla": True,
         "enable_weight_nz_layout": True
     }
-    speculative_config = {
-        "num_speculative_tokens": 1,
-        "method": "deepseek_mtp"
-    }
+    speculative_config = {"num_speculative_tokens": 1, "method": "mtp"}
     server_args = [
         "--quantization", "ascend", "--data-parallel-size", "2",
         "--tensor-parallel-size", "8", "--enable-expert-parallel", "--port",
diff --git a/tests/e2e/nightly/models/test_deepseek_r1_0528_w8a8.py b/tests/e2e/nightly/models/test_deepseek_r1_0528_w8a8.py
@@ -76,10 +76,7 @@ async def test_models(model: str, mode: str) -> None:
         "HCCL_BUFFSIZE": "1024",
         "PYTORCH_NPU_ALLOC_CONF": "expandable_segments:True"
     }
-    speculative_config = {
-        "num_speculative_tokens": 1,
-        "method": "deepseek_mtp"
-    }
+    speculative_config = {"num_speculative_tokens": 1, "method": "mtp"}
     additional_config = {
         "torchair_graph_config": {
             "enabled": True,
diff --git a/tests/e2e/nightly/models/test_deepseek_r1_w8a8_eplb.py b/tests/e2e/nightly/models/test_deepseek_r1_w8a8_eplb.py
@@ -62,10 +62,7 @@ async def test_models(model: str) -> None:
         "DISABLE_L2_CACHE": "1",
         "DYNAMIC_EPLB": "true",
     }
-    speculative_config = {
-        "num_speculative_tokens": 1,
-        "method": "deepseek_mtp"
-    }
+    speculative_config = {"num_speculative_tokens": 1, "method": "mtp"}
     compilation_config = {
         "cudagraph_capture_sizes": [24],
         "cudagraph_mode": "FULL_DECODE_ONLY"
diff --git a/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2-torchair.yaml b/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2-torchair.yaml
@@ -29,7 +29,7 @@ deployment:
       --trust-remote-code
       --quantization ascend
       --gpu-memory-utilization 0.9
-      --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
       --additional-config '{"torchair_graph_config":{"enabled":true,"enable_multistream_moe":true},"chunked_prefill_for_mla":true,"enable_weight_nz_layout":true}'
 
   -
@@ -50,7 +50,7 @@ deployment:
       --trust-remote-code
       --quantization ascend
       --gpu-memory-utilization 0.9
-      --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
       --additional-config '{"torchair_graph_config":{"enabled":true,"enable_multistream_moe":true},"chunked_prefill_for_mla":true,"enable_weight_nz_layout":true}'
 benchmarks:
   acc:
diff --git a/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2.yaml b/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-A2.yaml
@@ -30,7 +30,7 @@ deployment:
       --quantization ascend
       --gpu-memory-utilization 0.9
       --enforce-eager
-      --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
       --additional-config '{"chunked_prefill_for_mla":true,"enable_weight_nz_layout":true}'
 
   -
@@ -52,6 +52,6 @@ deployment:
       --quantization ascend
       --gpu-memory-utilization 0.9
       --enforce-eager
-      --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+      --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
       --additional-config '{"chunked_prefill_for_mla":true,"enable_weight_nz_layout":true}'
 benchmarks:
diff --git a/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-EPLB.yaml b/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8-EPLB.yaml
@@ -39,7 +39,7 @@ deployment:
           --max-num-batched-tokens 16384
           --trust-remote-code
           --gpu-memory-utilization 0.9
-          --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+          --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
           --kv-transfer-config
           '{"kv_connector": "LLMDataDistCMgrConnector",
           "kv_buffer_device": "npu",
@@ -69,7 +69,7 @@ deployment:
           --max-num-batched-tokens 16384
           --trust-remote-code
           --gpu-memory-utilization 0.9
-          --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+          --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
           --kv-transfer-config
           '{"kv_connector": "LLMDataDistCMgrConnector",
           "kv_buffer_device": "npu",
@@ -100,7 +100,7 @@ deployment:
         --max-num-batched-tokens 256
         --trust-remote-code
         --gpu-memory-utilization 0.9
-        --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+        --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
         --kv-transfer-config
         '{"kv_connector": "LLMDataDistCMgrConnector",
         "kv_buffer_device": "npu",
@@ -130,7 +130,7 @@ deployment:
         --max-num-batched-tokens 256
         --trust-remote-code
         --gpu-memory-utilization 0.9
-        --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+        --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
         --kv-transfer-config
         '{"kv_connector": "LLMDataDistCMgrConnector",
         "kv_buffer_device": "npu",
diff --git a/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8.yaml b/tests/e2e/nightly/multi_node/config/models/DeepSeek-R1-W8A8.yaml
@@ -38,7 +38,7 @@ deployment:
           --max-num-batched-tokens 16384
           --trust-remote-code
           --gpu-memory-utilization 0.9
-          --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+          --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
           --kv-transfer-config
           '{"kv_connector": "LLMDataDistCMgrConnector",
           "kv_buffer_device": "npu",
@@ -68,7 +68,7 @@ deployment:
           --max-num-batched-tokens 16384
           --trust-remote-code
           --gpu-memory-utilization 0.9
-          --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+          --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
           --kv-transfer-config
           '{"kv_connector": "LLMDataDistCMgrConnector",
           "kv_buffer_device": "npu",
@@ -99,7 +99,7 @@ deployment:
         --max-num-batched-tokens 256
         --trust-remote-code
         --gpu-memory-utilization 0.9
-        --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+        --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
         --kv-transfer-config
         '{"kv_connector": "LLMDataDistCMgrConnector",
         "kv_buffer_device": "npu",
@@ -129,7 +129,7 @@ deployment:
         --max-num-batched-tokens 256
         --trust-remote-code
         --gpu-memory-utilization 0.9
-        --speculative-config '{"num_speculative_tokens": 1, "method":"deepseek_mtp"}'
+        --speculative-config '{"num_speculative_tokens": 1, "method":"mtp"}'
         --kv-transfer-config
         '{"kv_connector": "LLMDataDistCMgrConnector",
         "kv_buffer_device": "npu",
diff --git a/tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py b/tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_correctness.py
@@ -56,7 +56,7 @@ def mtp_correctness(sampling_config: SamplingParams,
                     enable_expert_parallel=True,
                     speculative_config={
                         "method":
-                        "deepseek_mtp",
+                        "mtp",
                         "num_speculative_tokens":
                         num_speculative_tokens,
                         "disable_padded_drafter_batch":
diff --git a/tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_torchair_correctness.py b/tests/e2e/singlecard/spec_decode_v1/test_v1_mtp_torchair_correctness.py
@@ -58,7 +58,7 @@ def mtp_torchair_correctness(
                     distributed_executor_backend="mp",
                     enable_expert_parallel=True,
                     speculative_config={
-                        "method": "deepseek_mtp",
+                        "method": "mtp",
                         "num_speculative_tokens": 1,
                     },
                     enforce_eager=False,
diff --git a/tests/ut/torchair/test_torchair_model_runner.py b/tests/ut/torchair/test_torchair_model_runner.py
@@ -21,7 +21,7 @@ def setup_npu_torchair_model_runner(self, mocker: MockerFixture):
         runner.vllm_config = MagicMock(spec=VllmConfig)
 
         runner.speculative_config = MagicMock(
-            method="deepseek_mtp",
+            method="mtp",
             num_speculative_tokens=4,
             disable_padded_drafter_batch=False)
 
diff --git a/tests/ut/torchair/test_torchair_mtp_proposer.py b/tests/ut/torchair/test_torchair_mtp_proposer.py
@@ -19,7 +19,7 @@ def setup_torchair_mtp_proposer(self, mocker: MockerFixture):
         vllm_config.speculative_config = MagicMock()
         vllm_config.speculative_config.draft_model_config = MagicMock()
         vllm_config.speculative_config.draft_model_config.dtype = torch.float16
-        vllm_config.speculative_config.method = "deepseek_mtp"
+        vllm_config.speculative_config.method = "mtp"
         vllm_config.speculative_config.num_speculative_tokens = 5
         vllm_config.load_config = MagicMock()
         cache_config = CacheConfig(block_size=16)
diff --git a/vllm_ascend/compilation/acl_graph.py b/vllm_ascend/compilation/acl_graph.py
@@ -257,7 +257,7 @@ def update_mla_attn_params(update_stream, forward_context, runtime_shape,
              softmax_lse) = param
             seq_lens_list = forward_context.attn_metadata[
                 key].decode.seq_lens_list
-            if speculative_config and speculative_config.method == "deepseek_mtp" \
+            if speculative_config and speculative_config.method == "mtp" \
                     and not forward_context.is_mtp_model:
                 actual_seq_lengths = forward_context.attn_metadata[
                     key].decode.actual_seq_lengths_q
diff --git a/vllm_ascend/patch/platform/__init__.py b/vllm_ascend/patch/platform/__init__.py
@@ -16,7 +16,6 @@
 
 import os
 
-import vllm_ascend.patch.platform.patch_config  # noqa
 import vllm_ascend.patch.platform.patch_distributed  # noqa
 import vllm_ascend.patch.platform.patch_ec_connector  # noqa
 import vllm_ascend.patch.platform.patch_mamba_config  # noqa
diff --git a/vllm_ascend/patch/platform/patch_config.py b/vllm_ascend/patch/platform/patch_config.py
diff --git a/vllm_ascend/spec_decode/__init__.py b/vllm_ascend/spec_decode/__init__.py
diff --git a/vllm_ascend/torchair/torchair_mla.py b/vllm_ascend/torchair/torchair_mla.py
diff --git a/vllm_ascend/torchair/torchair_model_runner.py b/vllm_ascend/torchair/torchair_model_runner.py
diff --git a/vllm_ascend/torchair/torchair_sfa.py b/vllm_ascend/torchair/torchair_sfa.py
diff --git a/vllm_ascend/worker/model_runner_v1.py b/vllm_ascend/worker/model_runner_v1.py
