Disable profiling for AMD perf tests (keep profiling for NVIDIA)

python/sglang/test/nightly_utils.py

@@ -94,6 +94,7 @@ def build_benchmark_command(
         json_output_file: str,
         extra_args: Optional[List[str]] = None,
         server_args: Optional[List[str]] = None,
+        enable_profile: bool = True,
     ) -> List[str]:
         """Build the benchmark command with all required arguments.
 
@@ -106,6 +107,7 @@ def build_benchmark_command(
             json_output_file: Path to JSON output file
             extra_args: Optional extra arguments to append to command
             server_args: Optional server launch arguments to record in metrics
+            enable_profile: Whether to enable profiling (default True for NVIDIA)
 
         Returns:
             List of command arguments ready for subprocess.run()
@@ -125,15 +127,22 @@ def build_benchmark_command(
             "--output-len",
             *[str(x) for x in output_lens],
             "--show-report",
-            "--profile",
-            "--profile-by-stage",
-            "--profile-output-dir",
-            profile_path_prefix,
             f"--pydantic-result-filename={json_output_file}",
             "--no-append-to-github-summary",
             "--trust-remote-code",
         ]
 
+        # Add profiling flags only if enabled (disabled for AMD tests)
+        if enable_profile and profile_path_prefix:
+            command.extend(
+                [
+                    "--profile",
+                    "--profile-by-stage",
+                    "--profile-output-dir",
+                    profile_path_prefix,
+                ]
+            )
+
         if extra_args:
             command.extend(extra_args)
 
@@ -218,6 +227,7 @@ def run_benchmark_for_model(
         other_args: Optional[List[str]] = None,
         variant: str = "",
         extra_bench_args: Optional[List[str]] = None,
+        enable_profile: bool = True,
     ) -> Tuple[List[BenchmarkResult], bool, Optional[float]]:
         """Run a complete benchmark for a single model with server management.
 
@@ -236,6 +246,7 @@ def run_benchmark_for_model(
             other_args: Arguments to pass to server launch
             variant: Optional variant suffix (e.g., "basic", "mtp")
             extra_bench_args: Extra arguments for the benchmark command
+            enable_profile: Whether to enable profiling (default True for NVIDIA)
 
         Returns:
             Tuple of (list of BenchmarkResult objects, success_bool, avg_spec_accept_length or None)
@@ -273,6 +284,7 @@ def run_benchmark_for_model(
                 json_output_file,
                 extra_args=bench_args,
                 server_args=other_args,
+                enable_profile=enable_profile,
             )
 
             result, cmd_success = self.run_benchmark_command(command, model_description)

test/registered/amd/perf/mi30x/test_deepseek_v31_perf.py

@@ -129,6 +129,7 @@ def test_bench_one_batch(self):
                         other_args=variant_config["other_args"],
                         variant=variant_config["name"],
                         extra_bench_args=["--trust-remote-code"],
+                        enable_profile=False,  # Disable profiling for AMD tests
                     )
                     results = result_tuple[0]
                     success = result_tuple[1]

test/registered/amd/perf/mi30x/test_deepseek_v32_basic_perf_amd.py

@@ -115,6 +115,7 @@ def test_bench_one_batch(self):
                 other_args=self.variant_config["other_args"],
                 variant=self.variant_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi30x/test_deepseek_v32_mtp_perf_amd.py

@@ -124,6 +124,7 @@ def test_bench_one_batch(self):
                 other_args=self.variant_config["other_args"],
                 variant=self.variant_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi30x/test_deepseek_v3_perf.py

@@ -119,6 +119,7 @@ def test_bench_one_batch(self):
                         other_args=variant_config["other_args"],
                         variant=variant_config["name"],
                         extra_bench_args=["--trust-remote-code"],
+                        enable_profile=False,  # Disable profiling for AMD tests
                     )
                     results = result_tuple[0]
                     success = result_tuple[1]

test/registered/amd/perf/mi30x/test_grok1_fp8_perf.py

@@ -109,6 +109,7 @@ def test_bench_grok1_fp8(self):
                 other_args=self.model_config["other_args"],
                 variant=self.model_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi30x/test_grok1_int4_perf.py

@@ -119,6 +119,7 @@ def test_bench_grok1_int4(self):
                 other_args=self.model_config["other_args"],
                 variant=self.model_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi30x/test_grok2_perf.py

@@ -121,6 +121,7 @@ def test_bench_grok2(self):
                 other_args=self.model_config["other_args"],
                 variant=self.model_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi30x/test_text_models_perf_amd.py

@@ -110,6 +110,7 @@ def test_bench_one_batch(self):
                         input_lens=self.input_lens,
                         output_lens=self.output_lens,
                         other_args=other_args,
+                        enable_profile=False,  # Disable profiling for AMD tests
                     )
                     results = result_tuple[0]
                     success = result_tuple[1]

test/registered/amd/perf/mi30x/test_vlms_perf_amd.py

@@ -123,6 +123,7 @@ def test_bench_one_batch(self):
                         output_lens=self.output_lens,
                         other_args=other_args,
                         extra_bench_args=extra_bench_args,
+                        enable_profile=False,  # Disable profiling for AMD tests
                     )
                     results = result_tuple[0]
                     success = result_tuple[1]

test/registered/amd/perf/mi35x/test_deepseek_r1_mxfp4_perf_mi35x.py

@@ -152,6 +152,7 @@ def test_bench_one_batch(self):
                         other_args=variant_config["other_args"],
                         variant=variant_config["name"],
                         extra_bench_args=["--trust-remote-code"],
+                        enable_profile=False,  # Disable profiling for AMD tests
                     )
                     results = result_tuple[0]
                     success = result_tuple[1]

test/registered/amd/perf/mi35x/test_deepseek_v32_basic_perf_mi35x.py

@@ -114,6 +114,7 @@ def test_bench_one_batch(self):
                 other_args=self.variant_config["other_args"],
                 variant=self.variant_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi35x/test_deepseek_v32_mtp_perf_mi35x.py

@@ -96,6 +96,7 @@ def _run_benchmark_with_timeout(
             profile_path_prefix,
             json_output_file,
             extra_args=bench_args,
+            enable_profile=False,  # Disable profiling for AMD tests
         )
         _, cmd_success = runner.run_benchmark_command(command, model_description)
         if not cmd_success:

test/registered/amd/perf/mi35x/test_grok1_int4_perf_mi35x.py

@@ -112,6 +112,7 @@ def test_grok1_int4_perf(self):
                 other_args=self.model_config["other_args"],
                 variant=self.model_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]

test/registered/amd/perf/mi35x/test_grok2_perf_mi35x.py

@@ -112,6 +112,7 @@ def test_grok2_perf(self):
                 other_args=self.model_config["other_args"],
                 variant=self.model_config["name"],
                 extra_bench_args=["--trust-remote-code"],
+                enable_profile=False,  # Disable profiling for AMD tests
             )
             results = result_tuple[0]
             success = result_tuple[1]