[Cherry Pick] Two recent small diffs for transformers export

src/sparseml/pytorch/utils/exporter.py

@@ -460,7 +460,7 @@ def export_onnx(
         https://pytorch.org/docs/stable/onnx.html
     """
     if _PARSED_TORCH_VERSION >= version.parse("1.10.0") and opset < 13 and convert_qat:
-        warnings.warn(
+        raise ValueError(
             "Exporting onnx with QAT and opset < 13 may result in errors. "
             "Please use opset>=13 with QAT. "
             "See https://github.com/pytorch/pytorch/issues/77455 for more info. "

src/sparseml/transformers/export.py

@@ -336,8 +336,16 @@ def export_transformer_to_onnx(
         recipe_args=None,
         teacher=None,
     )
-
-    applied = trainer.apply_manager(epoch=math.inf, checkpoint=None)
+    try:
+        applied = trainer.apply_manager(epoch=math.inf, checkpoint=None)
+    except ValueError as e:
+        raise ValueError(
+            f"Failed to apply the recipe to the "
+            f"model with the exception message:\n{e}\n"
+            "It is possible, that there are missing modules "
+            "specific to the model, that were not properly loaded. "
+            "A possible solution would be setting the --trust_remote_code flag"
+        )
 
     if not applied:
         _LOGGER.warning(