code formatting

Signed-off-by: binghanc <176802681+binghanc@users.noreply.github.com>

Author: binghanc (generated by with_the_same_user script)

tensorrt_llm/serve/scripts/benchmark_serving.py

@@ -253,13 +253,20 @@ def calculate_metrics(
         percentiles_e2el_ms=[(p, np.percentile(e2els or 0, p) * 1000)
                              for p in selected_percentiles],
         tput_user=np.mean(tput_user or 0),
-        mean_avg_decoded_tokens_per_iter=np.mean(avg_decoded_tokens_per_iter_list or 0),
-        min_avg_decoded_tokens_per_iter=np.min(avg_decoded_tokens_per_iter_list) if avg_decoded_tokens_per_iter_list else 0.0,
-        max_avg_decoded_tokens_per_iter=np.max(avg_decoded_tokens_per_iter_list) if avg_decoded_tokens_per_iter_list else 0.0,
-        median_avg_decoded_tokens_per_iter=np.median(avg_decoded_tokens_per_iter_list or 0),
-        std_avg_decoded_tokens_per_iter=np.std(avg_decoded_tokens_per_iter_list or 0),
-        percentiles_avg_decoded_tokens_per_iter=[(p, np.percentile(avg_decoded_tokens_per_iter_list or 0, p))
-                                                  for p in selected_percentiles],
+        mean_avg_decoded_tokens_per_iter=np.mean(
+            avg_decoded_tokens_per_iter_list or 0),
+        min_avg_decoded_tokens_per_iter=np.min(avg_decoded_tokens_per_iter_list)
+        if avg_decoded_tokens_per_iter_list else 0.0,
+        max_avg_decoded_tokens_per_iter=np.max(avg_decoded_tokens_per_iter_list)
+        if avg_decoded_tokens_per_iter_list else 0.0,
+        median_avg_decoded_tokens_per_iter=np.median(
+            avg_decoded_tokens_per_iter_list or 0),
+        std_avg_decoded_tokens_per_iter=np.std(avg_decoded_tokens_per_iter_list
+                                               or 0),
+        percentiles_avg_decoded_tokens_per_iter=[
+            (p, np.percentile(avg_decoded_tokens_per_iter_list or 0, p))
+            for p in selected_percentiles
+        ],
     )
     return metrics, actual_output_lens
 
@@ -502,7 +509,10 @@ async def limited_request_func(request_func_input, streaming, pbar,
             "max": metrics.max_avg_decoded_tokens_per_iter,
             "median": metrics.median_avg_decoded_tokens_per_iter,
             "std": metrics.std_avg_decoded_tokens_per_iter,
-            "percentiles": {f"p{p}": v for p, v in metrics.percentiles_avg_decoded_tokens_per_iter}
+            "percentiles": {
+                f"p{p}": v
+                for p, v in metrics.percentiles_avg_decoded_tokens_per_iter
+            }
         },
         "input_lens": [output.prompt_len for output in outputs],
         "output_lens": actual_output_lens,
@@ -524,14 +534,15 @@ def process_one_metric(
     ):
         # This function prints and adds statistics of the specified metric.
         # Skip if not in selected metrics (except avg_decoded_tokens_per_iter which has its own condition)
-        if (metric_attribute_name not in selected_percentile_metrics and metric_attribute_name != "avg_decoded_tokens_per_iter"):
+        if (metric_attribute_name not in selected_percentile_metrics
+                and metric_attribute_name != "avg_decoded_tokens_per_iter"):
             return
-        
+
         # Build attribute suffix (e.g., "_ms" or "")
         attr_suffix = f"_{unit_suffix}" if unit_suffix else ""
         # Build display unit (e.g., " (ms)" or "")
         display_unit = f" ({unit_suffix})" if unit_suffix else ""
-        
+
         print("{s:{c}^{n}}".format(s=metric_header, n=50, c='-'))
         print("{:<40} {:<10.2f}".format(
             f"Mean {metric_name}{display_unit}:",
@@ -557,23 +568,25 @@ def process_one_metric(
                 getattr(metrics, f"max_{metric_attribute_name}{attr_suffix}")))
             result[f"max_{metric_attribute_name}{attr_suffix}"] = getattr(
                 metrics, f"max_{metric_attribute_name}{attr_suffix}")
-        
+
         result[f"mean_{metric_attribute_name}{attr_suffix}"] = getattr(
             metrics, f"mean_{metric_attribute_name}{attr_suffix}")
         result[f"median_{metric_attribute_name}{attr_suffix}"] = getattr(
             metrics, f"median_{metric_attribute_name}{attr_suffix}")
-        
-        for p, value in getattr(metrics,
-                                f"percentiles_{metric_attribute_name}{attr_suffix}"):
+
+        for p, value in getattr(
+                metrics, f"percentiles_{metric_attribute_name}{attr_suffix}"):
             p_word = str(int(p)) if int(p) == p else str(p)
-            print("{:<40} {:<10.2f}".format(f"P{p_word} {metric_name}{display_unit}:",
-                                            value))
+            print("{:<40} {:<10.2f}".format(
+                f"P{p_word} {metric_name}{display_unit}:", value))
             result[f"p{p_word}_{metric_attribute_name}{attr_suffix}"] = value
 
     # Print avg_decoded_tokens_per_iter statistics if available
     if metrics.mean_avg_decoded_tokens_per_iter > 0.0:
-        process_one_metric("avg_decoded_tokens_per_iter", "Avg Decoded Tokens per Iter",
-                          "Avg Decoded Tokens per Iter", unit_suffix="")
+        process_one_metric("avg_decoded_tokens_per_iter",
+                           "Avg Decoded Tokens per Iter",
+                           "Avg Decoded Tokens per Iter",
+                           unit_suffix="")
 
     process_one_metric("ttft", "TTFT", "Time to First Token")
     process_one_metric("tpot", "TPOT",