Fix live quick search (#513)

* Fix issue with not sweeping concurrencies

* Fix CPU vs GPU kind

* Fix issue with an error being printed out when it shouldn't

Author: tgerdesnv

model_analyzer/config/generate/quick_plus_concurrency_sweep_run_config_generator.py

@@ -87,8 +87,20 @@ def get_configs(self) -> Generator[RunConfig, None, None]:
         RunConfig
             The next RunConfig generated by this class
         """
+
+        logger.info("")
+        logger.info("Starting quick mode search to find optimal configs")
+        logger.info("")
         yield from self._execute_quick_search()
+        logger.info("")
+        logger.info(
+            "Done with quick mode search. Gathering concurrency sweep measurements for reports"
+        )
+        logger.info("")
         yield from self._sweep_concurrency_over_top_results()
+        logger.info("")
+        logger.info("Done gathering concurrency sweep measurements for reports")
+        logger.info("")
 
     def _execute_quick_search(self):
         self._rcg = self._create_quick_run_config_generator()

model_analyzer/config/generate/quick_run_config_generator.py

@@ -286,13 +286,14 @@ def _get_next_model_config(self, model_num: int) -> ModelConfig:
         dimension_values = self._get_coordinate_values(
             self._coordinate_to_measure, model_num)
 
+        kind = "KIND_CPU" if self._models[model_num].cpu_only() else "KIND_GPU"
         param_combo = {
             'dynamic_batching': {},
             'max_batch_size':
                 dimension_values['max_batch_size'],
             'instance_group': [{
                 'count': dimension_values['instance_count'],
-                'kind': "KIND_GPU",
+                'kind': kind,
             }]
         }
 

model_analyzer/result/result_manager.py

@@ -279,7 +279,7 @@ def _complete_profile_setup(self):
         else:
             self._setup_for_sequential_analysis()
 
-        self._add_results_to_heaps()
+        self._add_results_to_heaps(suppress_warnings=True)
 
     def _complete_analyze_setup(self):
         self._create_concurrent_analysis_model_name()
@@ -290,7 +290,7 @@ def _complete_analyze_setup(self):
             self._setup_for_sequential_analysis()
 
         self._check_for_models_in_checkpoint()
-        self._add_results_to_heaps()
+        self._add_results_to_heaps(suppress_warnings=False)
 
     def _create_concurrent_analysis_model_name(self):
         analysis_model_names = [
@@ -383,7 +383,7 @@ def _check_for_models_in_checkpoint(self):
                     f"The model {model_name} was not found in the loaded checkpoint."
                 )
 
-    def _add_results_to_heaps(self):
+    def _add_results_to_heaps(self, suppress_warnings=False):
         """
         Construct and add results to individual result heaps 
         as well as global result heap
@@ -392,7 +392,8 @@ def _add_results_to_heaps(self):
             'ResultManager.results')
 
         for model_name in self._analysis_model_names:
-            model_measurements = results.get_model_measurements_dict(model_name)
+            model_measurements = results.get_model_measurements_dict(
+                model_name, suppress_warnings)
 
             # Only add in models that exist in the checkpoint
             if not model_measurements: