Export API fix: Raise meaningful exception if model has no preprocessing metadata but preprocessing=True

src/super_gradients/module_interfaces/exportable_detector.py

@@ -21,7 +21,15 @@
 
 logger = get_logger(__name__)
 
-__all__ = ["ExportableObjectDetectionModel", "AbstractObjectDetectionDecodingModule", "ModelExportResult"]
+__all__ = ["ExportableObjectDetectionModel", "AbstractObjectDetectionDecodingModule", "ModelExportResult", "ModelHasNoPreprocessingParamsException"]
+
+
+class ModelHasNoPreprocessingParamsException(Exception):
+    """
+    Exception that is raised when model does not have preprocessing parameters.
+    """
+
+    pass
 
 
 class AbstractObjectDetectionDecodingModule(nn.Module):
@@ -295,7 +303,18 @@ def export(
         if isinstance(preprocessing, nn.Module):
             preprocessing_module = preprocessing
         elif preprocessing is True:
-            preprocessing_module = model.get_preprocessing_callback()
+            try:
+                preprocessing_module = model.get_preprocessing_callback()
+            except ModelHasNoPreprocessingParamsException:
+                raise ValueError(
+                    "It looks like your model does not have dataset preprocessing params properly set.\n"
+                    "This may happen if you instantiated model from scratch and not trained it yet. \n"
+                    "Here are what you can do to fix this:\n"
+                    "1. Manually fill up dataset processing params via model.set_dataset_processing_params(...).\n"
+                    "2. Train your model first and then export it. Trainer will set_dataset_processing_params(...) for you.\n"
+                    '3. Instantiate a model using pretrained weights: models.get(..., pretrained_weights="coco") \n'
+                    "4. Disable preprocessing by passing model.export(..., preprocessing=False). \n"
+                )
             if isinstance(preprocessing_module, nn.Sequential):
                 preprocessing_module = nn.Sequential(CastTensorTo(model_type), *iter(preprocessing_module))
             else:

src/super_gradients/training/models/detection_models/pp_yolo_e/pp_yolo_e.py

@@ -10,6 +10,7 @@
 from super_gradients.common.object_names import Models
 from super_gradients.common.registry.registry import register_model
 from super_gradients.module_interfaces import AbstractObjectDetectionDecodingModule, ExportableObjectDetectionModel, HasPredict
+from super_gradients.module_interfaces.exportable_detector import ModelHasNoPreprocessingParamsException
 from super_gradients.modules import RepVGGBlock
 from super_gradients.training.models.arch_params_factory import get_arch_params
 from super_gradients.training.models.detection_models.csp_resnet import CSPResNetBackbone
@@ -104,6 +105,8 @@ def get_post_prediction_callback(conf: float, iou: float) -> DetectionPostPredic
 
     def get_preprocessing_callback(self, **kwargs):
         processing = self.get_processing_params()
+        if processing is None:
+            raise ModelHasNoPreprocessingParamsException()
         preprocessing_module = processing.get_equivalent_photometric_module()
         return preprocessing_module
 

src/super_gradients/training/models/detection_models/yolo_base.py

@@ -11,6 +11,7 @@
 from super_gradients.common.decorators.factory_decorator import resolve_param
 from super_gradients.common.factories.processing_factory import ProcessingFactory
 from super_gradients.module_interfaces import ExportableObjectDetectionModel, AbstractObjectDetectionDecodingModule, HasPredict
+from super_gradients.module_interfaces.exportable_detector import ModelHasNoPreprocessingParamsException
 from super_gradients.modules import CrossModelSkipConnection, Conv
 from super_gradients.training.models.classification_models.regnet import AnyNetX, Stage
 from super_gradients.training.models.detection_models.csp_darknet53 import GroupedConvBlock, CSPDarknet53, get_yolo_type_params, SPP
@@ -502,6 +503,8 @@ def get_processing_params(self) -> Optional[Processing]:
 
     def get_preprocessing_callback(self, **kwargs):
         processing = self.get_processing_params()
+        if processing is None:
+            raise ModelHasNoPreprocessingParamsException()
         preprocessing_module = processing.get_equivalent_photometric_module()
         return preprocessing_module
 

src/super_gradients/training/models/detection_models/yolo_nas/yolo_nas_variants.py

@@ -8,7 +8,7 @@
 from super_gradients.common.abstractions.abstract_logger import get_logger
 from super_gradients.common.object_names import Models
 from super_gradients.common.registry import register_model
-from super_gradients.module_interfaces import ExportableObjectDetectionModel, AbstractObjectDetectionDecodingModule
+from super_gradients.module_interfaces import ExportableObjectDetectionModel, AbstractObjectDetectionDecodingModule, ModelHasNoPreprocessingParamsException
 from super_gradients.training.models.arch_params_factory import get_arch_params
 from super_gradients.training.models.detection_models.customizable_detector import CustomizableDetector
 from super_gradients.training.models.detection_models.pp_yolo_e import PPYoloEPostPredictionCallback
@@ -72,6 +72,8 @@ def get_decoding_module(self, num_pre_nms_predictions: int, **kwargs) -> Abstrac
 
     def get_preprocessing_callback(self, **kwargs):
         processing = self.get_processing_params()
+        if processing is None:
+            raise ModelHasNoPreprocessingParamsException()
         preprocessing_module = processing.get_equivalent_photometric_module()
         return preprocessing_module