feat(utils): rewrite segmentation benchmarker

Author: okunator

cellseg_models_pytorch/utils/latency_benchmark.py

@@ -12,7 +12,13 @@
 
 class LatencyBenchmarker:
     def __init__(self, inferer: BaseInferer) -> None:
-        """Benchmark latencies of the model an post-processing pipelines."""
+        """Benchmark latencies of the model an post-processing pipelines.
+
+        Parameters
+        ----------
+            inferer : BaseInferer
+                An inferer object that contains model outputs.
+        """
         self.inferer = inferer
 
         try:
@@ -178,7 +184,7 @@ def model_latency(
                 Height and width of the input.
             batch_size : int, default=1
                 Batch size.
-            reps : int, default=300
+            reps : int, default=100
                 Number of repetitions to run the latency measurement.
             warmup_reps : int, default=3
                 Number of repetitions that are used for warming up the gpu.