diff --git a/docs/tutorials/sca/tiny_aes.py b/docs/tutorials/sca/tiny_aes.py
index 4c0b2a25..1c57f3ae 100644
--- a/docs/tutorials/sca/tiny_aes.py
+++ b/docs/tutorials/sca/tiny_aes.py
@@ -197,6 +197,23 @@ def convert_to_sedpack(dataset_path: Path, original_files: Path) -> None:
     dataset.write_config()
 
 
+def process_batch(batch: dict[str, Any]) -> tuple[Any, dict[str, Any]]:
+    """Processing of a batch of records. The input is a dictionary of string
+    and tensor, the output of this function is a tuple the neural network's
+    input (trace) and a dictionary of one-hot encoded expected outputs.
+    """
+    # The first neural network was using just the first half of the trace:
+    inputs = batch["trace1"]
+    outputs = {
+        "sub_bytes_in_0":
+            keras.ops.one_hot(
+                batch["sub_bytes_in"][:, 0],
+                num_classes=256,
+            ),
+    }
+    return (inputs, outputs)
+
+
 def process_record(record: dict[str, Any]) -> tuple[Any, dict[str, Any]]:
     """Processing of a single record. The input is a dictionary of string and
     tensor, the output of this function is a tuple the neural network's input
@@ -256,20 +273,32 @@ def train(dataset_path: Path) -> None:
     )
     model.summary()
 
-    train_ds = dataset.as_tfdataset(
-        split="train",
-        process_record=process_record,
-        batch_size=batch_size,
-        #file_parallelism=4,
-        #parallelism=4,
-    )
-    validation_ds = dataset.as_tfdataset(
-        split="test",
-        process_record=process_record,
-        batch_size=batch_size,
-        #file_parallelism=4,
-        #parallelism=4,
-    )
+    match keras.backend.backend():
+        case "tensorflow":
+            train_ds = dataset.as_tfdataset(
+                split="train",
+                process_record=process_record,
+                batch_size=batch_size,
+            )
+            validation_ds = dataset.as_tfdataset(
+                split="test",
+                process_record=process_record,
+                batch_size=batch_size,
+            )
+        case "jax" | "torch":
+            train_ds = dataset.as_numpy_iterator_rust_batched(
+                split="train",
+                process_batch=process_batch,
+                batch_size=batch_size,
+            )
+            validation_ds = dataset.as_numpy_iterator_rust_batched(
+                split="test",
+                process_batch=process_batch,
+                batch_size=batch_size,
+            )
+        case _:
+            print(f"TODO support {keras.backend.backend() = }")
+            return
 
     # Train the model.
     _ = model.fit(