Review comments and unit fixed

triton-inference-server · Oct 14, 2024 · b7986bd · b7986bd
1 parent c34d91b
commit b7986bd
Show file tree

Hide file tree

Showing 3 changed files with 2 additions and 5 deletions.
diff --git a/src/metric_model_reporter.h b/src/metric_model_reporter.h
@@ -87,7 +87,7 @@ class MetricModelReporter {
   const MetricReporterConfig& Config();
   // Lookup counter metric by name, and increment it by value if it exists.
   void IncrementCounter(const std::string& name, double value);
-  // Increase gauge by value.
+  // Set gauge value.
   void SetGauge(const std::string& name, double value);
   // Increase gauge by value.
   void IncrementGauge(const std::string& name, double value);

diff --git a/src/metrics.cc b/src/metrics.cc
@@ -111,7 +111,7 @@ Metrics::Metrics()
 
       model_load_time_family_(prometheus::BuildGauge()
                                   .Name("nv_model_load_time")
-                                  .Help("Load Time per-model in nanoseconds")
+                                  .Help("Load Time per-model in seconds")
                                   .Register(*registry_)),
 
       pinned_memory_pool_total_family_(

diff --git a/src/model_repository_manager/model_lifecycle.cc b/src/model_repository_manager/model_lifecycle.cc
@@ -528,7 +528,6 @@ ModelLifeCycle::AsyncLoad(
     load_pool_->Enqueue([this, model_id, version, model_info, OnComplete,
                          load_tracker, is_config_provided]() {
       for (size_t retry = 0; retry <= options_.load_retry; ++retry) {
-        // TODO add here
         model_info->state_ = ModelReadyState::LOADING;
         CreateModel(model_id, version, model_info, is_config_provided);
         // Model state will be changed to NOT loading if failed to load,
@@ -866,8 +865,6 @@ ModelLifeCycle::ReportModelLoadTime(
     const std::chrono::duration<double>& time_to_load)
 {
 #ifdef TRITON_ENABLE_METRICS
-  // Pending request count should always be 0 or 1 per-request. A request should
-  // not decrement the count unless it has already been incremented.
   if (reporter) {
     double load_time_in_seconds = time_to_load.count();
     reporter->SetGauge(kModelLoadTimeMetric, load_time_in_seconds);