Classify authoring issues with model arch as input errors (#3837)

Summary:
Pull Request resolved: #3837

Sometimes users have mistakes in model arch, where an expected prediction / weight name is not present in model output when torchrec expects it to be there.
torchrec is open source code, so we can't change that assertion error to throw TrainingPlatforUserError. However, we can change it to raise a value error with a recognizable regex pattern so we can re-classify it in our logger as input error

Differential Revision: D95415280

fbshipit-source-id: 50aa80e5092d5ed67918b675847614d89df23622

Author: derekwan1

torchrec/metrics/model_utils.py

@@ -61,8 +61,12 @@ def parse_model_outputs(
         assert not weight_name, "weight name must be empty if prediction name is empty"
         return (labels, None, None)
     assert isinstance(labels, torch.Tensor)
+    if prediction_name not in model_out:
+        raise ValueError(f"Prediction name {prediction_name} not found in model output")
     predictions = model_out[prediction_name].squeeze()
     assert isinstance(predictions, torch.Tensor)
+    if weight_name not in model_out:
+        raise ValueError(f"Weight {weight_name} not found in model output")
     weights = model_out[weight_name].squeeze()
     assert isinstance(weights, torch.Tensor)
 
@@ -81,25 +85,25 @@ def parse_model_outputs(
             if is_vector_valued_label_and_prediction:
                 logger.warning(
                     f"""
-                    Vector valued labels and predictions are provided. 
+                    Vector valued labels and predictions are provided.
 
-                    For vector valued label and prediction we should have shapes 
+                    For vector valued label and prediction we should have shapes
                     labels.shape: (batch_size, dim_vector_valued_label)
                     predictions.shape: (batch_size, dim_vector_valued_prediction)
                     weights.shape: (batch_size,)
 
-                    The provided labels, predictions and weights comply with the conditions for vector valued labels and predictions. 
-                    These conditions are: 
+                    The provided labels, predictions and weights comply with the conditions for vector valued labels and predictions.
+                    These conditions are:
                     1. labels.dim() == 2
                     2. predictions.dim() == 2
                     3. weights.dim() == 1
                     4. labels.size()[0] == predictions.size()[0]
                     5. labels.size()[0] == weights.size()[0]
 
-                    The shapes of labels, predictions and weights are: 
-                    labels.shape == {labels.shape}, 
-                    predictions.shape == {predictions.shape}, 
-                    weights.shape == {weights.shape} 
+                    The shapes of labels, predictions and weights are:
+                    labels.shape == {labels.shape},
+                    predictions.shape == {predictions.shape},
+                    weights.shape == {weights.shape}
                     """
                 )
             else: