revert epoch to steps back, change templates

openvinotoolkit · Nov 29, 2024 · de0c0cb · de0c0cb
1 parent 3d8a4b4
commit de0c0cb
Show file tree

Hide file tree

Showing 59 changed files with 81 additions and 222 deletions.
diff --git a/src/otx/core/schedulers/warmup_schedulers.py b/src/otx/core/schedulers/warmup_schedulers.py
@@ -29,7 +29,7 @@ def __init__(
         self,
         optimizer: Optimizer,
         num_warmup_steps: int = 1000,
-        interval: Literal["step", "epoch"] = "epoch",
+        interval: Literal["step", "epoch"] = "step",
     ):
         if not num_warmup_steps > 0:
             msg = f"num_warmup_steps should be > 0, got {num_warmup_steps}"
@@ -65,7 +65,7 @@ def __init__(
         self,
         main_scheduler_callable: LRSchedulerCallable,
         num_warmup_steps: int = 0,
-        warmup_interval: Literal["step", "epoch"] = "epoch",
+        warmup_interval: Literal["step", "epoch"] = "step",
         monitor: str | None = None,
     ):
         self.main_scheduler_callable = SchedulerCallableSupportHPO.from_callable(main_scheduler_callable)

diff --git a/src/otx/recipe/detection/atss_mobilenetv2.yaml b/src/otx/recipe/detection/atss_mobilenetv2.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/atss_mobilenetv2_tile.yaml b/src/otx/recipe/detection/atss_mobilenetv2_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/atss_resnext101.yaml b/src/otx/recipe/detection/atss_resnext101.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/atss_resnext101_tile.yaml b/src/otx/recipe/detection/atss_resnext101_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/rtdetr_101_tile.yaml b/src/otx/recipe/detection/rtdetr_101_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 5
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/rtdetr_18_tile.yaml b/src/otx/recipe/detection/rtdetr_18_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 5
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/rtdetr_50_tile.yaml b/src/otx/recipe/detection/rtdetr_50_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 5
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/rtmdet_tiny.yaml b/src/otx/recipe/detection/rtmdet_tiny.yaml
@@ -12,7 +12,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/rtmdet_tiny_tile.yaml b/src/otx/recipe/detection/rtmdet_tiny_tile.yaml
@@ -12,7 +12,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/ssd_mobilenetv2.yaml b/src/otx/recipe/detection/ssd_mobilenetv2.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/ssd_mobilenetv2_tile.yaml b/src/otx/recipe/detection/ssd_mobilenetv2_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_l.yaml b/src/otx/recipe/detection/yolox_l.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_l_tile.yaml b/src/otx/recipe/detection/yolox_l_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_s.yaml b/src/otx/recipe/detection/yolox_s.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_s_tile.yaml b/src/otx/recipe/detection/yolox_s_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_tiny.yaml b/src/otx/recipe/detection/yolox_tiny.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_tiny_tile.yaml b/src/otx/recipe/detection/yolox_tiny_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_x.yaml b/src/otx/recipe/detection/yolox_x.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/detection/yolox_x_tile.yaml b/src/otx/recipe/detection/yolox_x_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 3
+        num_warmup_steps: 0
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_efficientnetb2b.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_efficientnetb2b.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_efficientnetb2b_tile.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_efficientnetb2b_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_r50.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_r50.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_r50_tile.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_r50_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_r50_tv.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_r50_tv.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_r50_tv_tile.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_r50_tv_tile.yaml
@@ -13,7 +13,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/instance_segmentation/maskrcnn_swint.yaml b/src/otx/recipe/instance_segmentation/maskrcnn_swint.yaml
@@ -12,7 +12,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/semantic_segmentation/litehrnet_18.yaml b/src/otx/recipe/semantic_segmentation/litehrnet_18.yaml
@@ -16,7 +16,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/semantic_segmentation/litehrnet_s.yaml b/src/otx/recipe/semantic_segmentation/litehrnet_s.yaml
@@ -16,7 +16,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/recipe/semantic_segmentation/litehrnet_x.yaml b/src/otx/recipe/semantic_segmentation/litehrnet_x.yaml
@@ -16,7 +16,7 @@ model:
     scheduler:
       class_path: otx.core.schedulers.LinearWarmupSchedulerCallable
       init_args:
-        num_warmup_steps: 10
+        num_warmup_steps: 100
         main_scheduler_callable:
           class_path: lightning.pytorch.cli.ReduceLROnPlateau
           init_args:

diff --git a/src/otx/tools/templates/classification/configuration.yaml b/src/otx/tools/templates/classification/configuration.yaml
@@ -87,11 +87,11 @@ learning_parameters:
   num_iters:
     affects_outcome_of: TRAINING
     default_value: 200
-    description:
-      Increasing this value causes the results to be more robust but training
-      time will be longer.
+    description: Maximum number of epochs to train a model.
+      Increasing this value may result in longer training, but potentially in a more robust model.
+      Note, if the early stopping is enabled, the actual number of epochs may be less than this value.
     editable: true
-    header: Number of training iterations
+    header: Number of training epochs
     max_value: 1000
     min_value: 1
     type: INTEGER
@@ -109,9 +109,9 @@ learning_parameters:
     description:
       Increasing this value might improve training speed however it might
       cause out of memory errors. If the number of workers is set to zero, data loading
-      will happen in the main training thread.
+      will happen only in the main training thread leading to decreasing training speed.
     editable: true
-    header: Number of cpu threads to use during batch generation
+    header: Number of cpu threads for batch generation (num workers)
     max_value: 8
     min_value: 0
     type: INTEGER
@@ -191,21 +191,19 @@ learning_parameters:
     warning: This is applied exclusively when early stopping is enabled.
   early_stop_iteration_patience:
     affects_outcome_of: TRAINING
-    default_value: 0
-    description:
-      Training will stop if the model does not improve within the number of iterations of patience.
-      This ensures the model is trained enough with the number of iterations of patience before early stopping.
+    default_value: 30
+    description: This parameter ensures that the model is trained enough within the number of iterations of warmup and stable before early stopping is on.
     editable: true
-    header: Iteration patience for early stopping
-    max_value: 1000
+    header: Number of warmup iterations for early stopping
+    max_value: 10000
     min_value: 0
     type: INTEGER
     ui_rules:
       action: DISABLE_EDITING
       operator: AND
       rules: []
       type: UI_RULES
-    value: 0
+    value: 30
     visible_in_ui: true
     warning: This is applied exclusively when early stopping is enabled.
   use_adaptive_interval:
@@ -264,7 +262,7 @@ learning_parameters:
   auto_num_workers:
     affects_outcome_of: TRAINING
     default_value: false
-    description: Adapt num_workers according to current hardware status automatically.
+    description: Adapt number of workers according to current hardware status automatically.
     editable: true
     header: Enable auto adaptive num_workers
     type: BOOLEAN