Update [guides][core]evaluation.ipynb

docs/notebooks/[guides][core]evaluation.ipynb

@@ -11,11 +11,9 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "In contrast to some machine learning frameworks that handle different behaviors of layers during training and evaluation with specific methods like `model.eval()` and `model.fit()`, `serket` offers a more explicit approach for managing this behavior. `serket` provides a function called `tree_eval`, which allows users to modify a tree of layers to disable any training-related behavior. The significant difference is that instead of modifying the layers in place `serket` replaces them with their evaluation counterparts.\n",
+    "Serket uses dispatching over layers to determine the evaluation counterpart of a layer.For instance, during evaluation, the `Dropout` layer is replaced by an `Identity` layer, and the `BatchNorm` layer is replaced by an `EvalBatchNorm` layer. This ensures that when evaluating a tree of layers, users can follow the **\"What you see is what you get\" (WYSIWYG)** principle, meaning that Seeing a `Dropout`/`BatchNorm` in the model have one meaning.\n",
     "\n",
-    "For instance, during evaluation, the `Dropout` layer is replaced by an `Identity` layer, and the `BatchNorm` layer is replaced by an `EvalBatchNorm` layer. This ensures that when evaluating a tree of layers, users can follow the **\"What you see is what you get\" (WYSIWYG)** principle, meaning that Seeing a `Dropout`/`BatchNorm` in the model have one meaning.\n",
-    "\n",
-    "Moreover, when building neural networks that includes the likes of `Dropout`, By design threading a training/eval flag is no longer required."
+    " The dispatch design choice allow user to define layers like `Dropout` and `BatchNorm` that have different behaviors during training and evaluation without having to threading a training/eval flag through the layers."
    ]
   },
   {