[CI] Append test variant name to markdown report header in nightly test (#16166)

python/sglang/test/nightly_bench_utils.py

@@ -87,7 +87,9 @@ Note: To view the traces through perfetto-ui, please:
         return f"| {self.batch_size} | {self.input_len} | {self.latency:.2f} | {self.input_throughput:.2f} | {self.output_throughput:.2f} | {accept_length} | {itl:.2f} | {input_cost:.2f} | {output_cost:.2f} | {profile_link} |\n"
 
 
-def generate_markdown_report(trace_dir, results: List[BenchmarkResult]) -> str:
+def generate_markdown_report(
+    trace_dir, results: List[BenchmarkResult], variant: Optional[str] = None
+) -> str:
     """Generate a markdown report from a list of BenchmarkResult object from a single run."""
     # Build model header with run_name if it's not "default"
     model_header = results[0].model_path
@@ -99,6 +101,9 @@ def generate_markdown_report(trace_dir, results: List[BenchmarkResult]) -> str:
     if gpu_config:
         model_header += f" [{gpu_config}]"
 
+    if variant:
+        model_header += f" ({variant})"
+
     summary = f"### {model_header}\n"
 
     summary += "| batch size | input len | latency (s) | input throughput (tok/s)  | output throughput (tok/s) | acc length | ITL (ms) | input cost ($/1M) | output cost ($/1M) | profile (extend) | profile (decode)|\n"

python/sglang/test/nightly_utils.py

@@ -309,14 +309,16 @@ class NightlyBenchmarkRunner:
             print(f"  Warning: Could not fetch spec accept length: {e}")
         return None
 
-    def add_report(self, results: List[BenchmarkResult]) -> None:
+    def add_report(
+        self, results: List[BenchmarkResult], variant: Optional[str] = None
+    ) -> None:
         """Add benchmark results to the full report.
 
         Args:
             results: List of BenchmarkResult objects to add to report
         """
         if results:
-            report_part = generate_markdown_report(self.profile_dir, results)
+            report_part = generate_markdown_report(self.profile_dir, results, variant)
             self.full_report += report_part + "\n"
 
     def write_final_report(self) -> None:

python/sglang/test/performance_test_runner.py

@@ -57,6 +57,7 @@ def run_performance_test(
 
     print(f"\n{'='*60}")
     print(f"Running PERFORMANCE test for {model.model_path}")
+    print(f"  Variant: {model.variant}")
     print(f"  Batch sizes: {batch_sizes}")
     print(f"  Input lens: {input_lens}")
     print(f"  Output lens: {output_lens}")
@@ -80,7 +81,7 @@ def run_performance_test(
         )
 
         if success and results:
-            perf_runner.add_report(results)
+            perf_runner.add_report(results, variant=model.variant)
             print(f"✓ Performance test succeeded for {model.model_path}")
 
             # Validate speculative decoding accept length if threshold is set

python/sglang/test/test_utils.py

@@ -1767,11 +1767,13 @@ class ModelLaunchSettings:
         tp_size: int = 1,
         extra_args: Optional[List[str]] = None,
         env: Optional[dict] = None,
+        variant: Optional[str] = None,
     ):
         self.model_path = model_path
         self.tp_size = tp_size
         self.extra_args = list(extra_args) if extra_args else []
         self.env = env
+        self.variant = variant
 
         if self.tp_size > 1 and "--tp" not in self.extra_args:
             self.extra_args.extend(["--tp", str(self.tp_size)])