Use spec v2 by default (#21062)

test/manual/ascend/test_ascend_deepseek_mtp.py

@@ -53,7 +53,6 @@ class TestAscendDeepSeekMTP(CustomTestCase):
         ]
 
         envs.SGLANG_NPU_USE_MLAPO.set(True)
-        envs.SGLANG_ENABLE_SPEC_V2.set(True)
         envs.SGLANG_ENABLE_OVERLAP_PLAN_STREAM.set(True)
 
     def test_a_gsm8k(self):

test/manual/test_deepseek_v31.py

@@ -50,7 +50,6 @@ class TestDeepseekV31(unittest.TestCase):
                 DEEPSEEK_V31_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + mtp_args,
-                env={"SGLANG_ENABLE_SPEC_V2": "1"},
                 variant="TP8+MTP",
             ),
         ]

test/manual/test_glm_46_fp8.py

@@ -41,7 +41,6 @@ class TestGLM46FP8(unittest.TestCase):
                 GLM_4_6_FP8_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + mtp_args,
-                env={"SGLANG_ENABLE_SPEC_V2": "1"},
                 variant="TP8+MTP",
             ),
         ]

test/manual/test_qwen3_235b.py

@@ -52,7 +52,6 @@ class TestQwen3235BFP8(unittest.TestCase):
                 QWEN3_235B_FP8_MODEL_PATH,
                 tp_size=8,
                 extra_args=base_args + eagle3_args,
-                env={"SGLANG_ENABLE_SPEC_V2": "1"},
                 variant="TP8+EP2+EAGLE3",
             ),
         ]