From 9819cf178397de4d47dc321434cb5da57adc7168 Mon Sep 17 00:00:00 2001
From: unknown <ppalanji@usc.edu>
Date: Wed, 22 Jan 2025 18:31:53 -0800
Subject: [PATCH 1/2] Added early stopping & patience parameter, needs cleanup

---
 detectron2/engine/hooks.py | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/detectron2/engine/hooks.py b/detectron2/engine/hooks.py
index fc37af0fd3..a14da69ce1 100644
--- a/detectron2/engine/hooks.py
+++ b/detectron2/engine/hooks.py
@@ -220,6 +220,7 @@ def __init__(
         eval_period: int,
         checkpointer: Checkpointer,
         val_metric: str,
+        patience: int, # Added by Peter
         mode: str = "max",
         file_prefix: str = "model_best",
     ) -> None:
@@ -235,6 +236,7 @@ def __init__(
         self._logger = logging.getLogger(__name__)
         self._period = eval_period
         self._val_metric = val_metric
+        self._patience = patience # Added by Peter
         assert mode in [
             "max",
             "min",
@@ -297,6 +299,19 @@ def after_step(self):
             and next_iter != self.trainer.max_iter
         ):
             self._best_checking()
+            
+            # Early stopping code added by Peter            
+            if (self.best_iter is None): 
+                self._logger.warning("Best iteration is None. This is likely due to the metric not being computed yet.")
+                return
+            iterations_without_improvement = (self.trainer.iter-self.best_iter) // self._period
+            if(iterations_without_improvement == 0):
+                return 
+            
+            self._logger.warning(f"Model has not improved for {iterations_without_improvement} evaluation cycles. RAW={self.trainer.iter-self.best_iter}Max is {self._patience}") 
+            if(iterations_without_improvement > self._patience): 
+                self._logger.warning(f"Early stopping at iteration: {self.trainer.iter}")
+                raise Exception("Early stopping exception. Terminating 'gracefully'")
 
     def after_train(self):
         # same conditions as `EvalHook`

From 3ccb191685371f2007e8e5f2af33ee9813bda87c Mon Sep 17 00:00:00 2001
From: unknown <ppalanji@usc.edu>
Date: Thu, 23 Jan 2025 11:49:05 -0800
Subject: [PATCH 2/2] Code cleanup

---
 detectron2/engine/hooks.py | 28 +++++++++++++++++-----------
 1 file changed, 17 insertions(+), 11 deletions(-)

diff --git a/detectron2/engine/hooks.py b/detectron2/engine/hooks.py
index a14da69ce1..26bc045d9f 100644
--- a/detectron2/engine/hooks.py
+++ b/detectron2/engine/hooks.py
@@ -220,9 +220,9 @@ def __init__(
         eval_period: int,
         checkpointer: Checkpointer,
         val_metric: str,
-        patience: int, # Added by Peter
         mode: str = "max",
         file_prefix: str = "model_best",
+        patience: int = None,
     ) -> None:
         """
         Args:
@@ -232,11 +232,12 @@ def __init__(
             mode (str): one of {'max', 'min'}. controls whether the chosen val metric should be
                 maximized or minimized, e.g. for "bbox/AP50" it should be "max"
             file_prefix (str): the prefix of checkpoint's filename, defaults to "model_best"
+            patience (int): the number of evaluation cycles without improvement before early stopping
         """
         self._logger = logging.getLogger(__name__)
         self._period = eval_period
         self._val_metric = val_metric
-        self._patience = patience # Added by Peter
+        self._patience = patience 
         assert mode in [
             "max",
             "min",
@@ -300,18 +301,23 @@ def after_step(self):
         ):
             self._best_checking()
             
-            # Early stopping code added by Peter            
-            if (self.best_iter is None): 
-                self._logger.warning("Best iteration is None. This is likely due to the metric not being computed yet.")
+            if self._patience is None or self.best_iter is None:
                 return
-            iterations_without_improvement = (self.trainer.iter-self.best_iter) // self._period
-            if(iterations_without_improvement == 0):
-                return 
             
-            self._logger.warning(f"Model has not improved for {iterations_without_improvement} evaluation cycles. RAW={self.trainer.iter-self.best_iter}Max is {self._patience}") 
+            iterations_without_improvement = (self.trainer.iter-self.best_iter) // self._period
+
             if(iterations_without_improvement > self._patience): 
-                self._logger.warning(f"Early stopping at iteration: {self.trainer.iter}")
-                raise Exception("Early stopping exception. Terminating 'gracefully'")
+                self._logger.info(
+                    f"Early stopping triggered at iteration {self.trainer.iter} due to lack of improvement " 
+                    f"after {iterations_without_improvement} cycles."
+                )                
+                raise Exception("Early stopping triggered. Terminating training process.")
+
+            if(iterations_without_improvement > 0):
+                self._logger.info(
+                    f"No improvement detected in the last {iterations_without_improvement} evaluation cycles. " 
+                    f"{self._patience - iterations_without_improvement} cycles remain before early stopping."
+                )
 
     def after_train(self):
         # same conditions as `EvalHook`