feat: add save best model mechanism

fastreid/engine/hooks.py

@@ -17,7 +17,7 @@ from fastreid.evaluation.testing import flatten_results_dict
 from fastreid.solver import optim
 from fastreid.utils import comm
 from fastreid.utils.checkpoint import PeriodicCheckpointer as _PeriodicCheckpointer
-from fastreid.utils.events import EventStorage, EventWriter
+from fastreid.utils.events import EventStorage, EventWriter, get_event_storage
 from fastreid.utils.file_io import PathManager
 from fastreid.utils.precision_bn import update_bn_stats, get_bn_modules
 from fastreid.utils.timer import Timer
@@ -198,10 +198,18 @@ class PeriodicCheckpointer(_PeriodicCheckpointer, HookBase):
 
     def before_train(self):
         self.max_epoch = self.trainer.max_epoch
+        if len(self.trainer.cfg.DATASETS.TESTS) == 1:
+            self.metric_name = "metric"
+        else:
+            self.metric_name = self.trainer.cfg.DATASETS.TESTS[0] + "/metric"
 
     def after_epoch(self):
         # No way to use **kwargs
-        self.step(self.trainer.epoch)
+        storage = get_event_storage()
+        metric_dict = dict(
+            metric=storage.latest()[self.metric_name][0] if self.metric_name in storage.latest() else -1
+        )
+        self.step(self.trainer.epoch, **metric_dict)
 
 
 class LRScheduler(HookBase):
@@ -473,7 +481,7 @@ class LayerFreeze(HookBase):
                     p.requires_grad_(False)
 
         self.is_frozen = True
-        freeze_layers = ",".join(self.freeze_layers)
+        freeze_layers = ", ".join(self.freeze_layers)
         self._logger.info(f'Freeze layer group "{freeze_layers}" training for {self.freeze_iters:d} iterations')
 
     def open_all_layer(self):

fastreid/evaluation/reid_evaluation.py

@@ -106,6 +106,7 @@ class ReidEvaluator(DatasetEvaluator):
             self._results['Rank-{}'.format(r)] = cmc[r - 1]
         self._results['mAP'] = mAP
         self._results['mINP'] = mINP
+        self._results["metric"] = (mAP + cmc[0]) / 2
 
         if self.cfg.TEST.ROC_ENABLED:
             scores, labels = evaluate_roc(dist, query_pids, gallery_pids, query_camids, gallery_camids)