[diffusion] feat: add performance mode server args (#24491)

test/registered/spec/dflash/test_dflash.py

@@ -53,12 +53,10 @@ class TestDFlashServerBase(CustomTestCase, MatchedStopMixin, GSM8KMixin):
         old_value = os.environ.get("SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN")
         os.environ["SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN"] = "1"
         try:
-            with envs.SGLANG_ENABLE_STRICT_MEM_CHECK_DURING_BUSY.override(
-                1
-            ), envs.SGLANG_SPEC_NAN_DETECTION.override(
-                True
-            ), envs.SGLANG_SPEC_OOB_DETECTION.override(
-                True
+            with (
+                envs.SGLANG_ENABLE_STRICT_MEM_CHECK_DURING_BUSY.override(1),
+                envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+                envs.SGLANG_SPEC_OOB_DETECTION.override(True),
             ):
                 cls.process = popen_launch_server(
                     cls.model,

test/registered/spec/eagle/test_deepseek_v3_fp4_mtp_small.py

@@ -49,9 +49,10 @@ class TestDeepseekV3FP4MTP(CustomTestCase):
             "--model-loader-extra-config",
             '{"enable_multithread_load": true,"num_threads": 64}',
         ]
-        with envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(True):
+        with (
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+        ):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,

test/registered/spec/eagle/test_eagle_constrained_decoding.py

@@ -59,12 +59,10 @@ class TestEagleConstrainedDecoding(
             cls.grammar_backend,
         ]
         launch_args.extend(cls.other_launch_args)
-        with envs.SGLANG_ENABLE_SPEC_V2.override(
-            cls.spec_v2
-        ), envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(
-            True
+        with (
+            envs.SGLANG_ENABLE_SPEC_V2.override(cls.spec_v2),
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
         ):
             cls.process = popen_launch_server(
                 cls.model,

test/registered/spec/eagle/test_eagle_dp_attention.py

@@ -57,9 +57,10 @@ class TestEAGLE3EngineDPAttention(CustomTestCase):
             "--cuda-graph-max-bs",
             "64",
         ]
-        with envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(True):
+        with (
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+        ):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,

test/registered/spec/eagle/test_eagle_infer_beta.py

@@ -63,14 +63,11 @@ class TestEagle3ServerBase(CustomTestCase, MatchedStopMixin):
             *[str(i) for i in range(1, cls.max_running_requests + 1)],
         ]
         launch_args.extend(cls.other_launch_args)
-        with envs.SGLANG_ENABLE_STRICT_MEM_CHECK_DURING_BUSY.override(
-            1
-        ), envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(
-            True
-        ), envs.SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN.override(
-            True
+        with (
+            envs.SGLANG_ENABLE_STRICT_MEM_CHECK_DURING_BUSY.override(1),
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+            envs.SGLANG_ALLOW_OVERWRITE_LONGER_CONTEXT_LEN.override(True),
         ):
             cls.process = popen_launch_server(
                 cls.model,

test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py

@@ -65,9 +65,10 @@ class TestEagleDPAttnServerSmall(CustomTestCase):
             "--speculative-num-draft-tokens",
             "4",
         ]
-        with envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(True):
+        with (
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+        ):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,

test/registered/spec/eagle/test_eagle_infer_beta_dp_attention_large.py

@@ -73,9 +73,10 @@ class TestEagleDPAttnServerLarge(CustomTestCase):
             "--model-loader-extra-config",
             '{"enable_multithread_load": true,"num_threads": 64}',
         ]
-        with envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(True):
+        with (
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+        ):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,

test/registered/spec/test_constrained_decoding_spec_reasoning.py

@@ -51,9 +51,10 @@ class ServerWithGrammar(CustomTestCase):
             "--speculative-num-draft-tokens=8",
         ]
 
-        with envs.SGLANG_SPEC_NAN_DETECTION.override(
-            True
-        ), envs.SGLANG_SPEC_OOB_DETECTION.override(True):
+        with (
+            envs.SGLANG_SPEC_NAN_DETECTION.override(True),
+            envs.SGLANG_SPEC_OOB_DETECTION.override(True),
+        ):
             cls.process = popen_launch_server(
                 cls.model,
                 cls.base_url,