[AIR][Docs] Clarify how LGBM/XGB trainers work (#28122)

doc/source/train/gbdt.rst

@@ -36,8 +36,18 @@ Ray-specific params are passed in through the trainer constructors.
 How to scale out training?
 --------------------------
 The benefit of using Ray AIR is that you can seamlessly scale up your training by
-adjusting the :class:`ScalingConfig <ray.air.config.ScalingConfig>`. Here are some
-examples for common use-cases:
+adjusting the :class:`ScalingConfig <ray.air.config.ScalingConfig>`.
+
+.. note::
+    Ray Train does not modify or otherwise alter the working
+    of the underlying XGBoost / LightGBM distributed training algorithms.
+    Ray only provides orchestration, data ingest and fault tolerance.
+    For more information on GBDT distributed training, refer to
+    `XGBoost documentation <https://xgboost.readthedocs.io>`__ and
+    `LightGBM documentation <https://lightgbm.readthedocs.io/>`__.
+
+
+Here are some examples for common use-cases:
 
 
 .. tabbed:: Multi-node CPU

python/ray/train/lightgbm/lightgbm_trainer.py

@@ -24,6 +24,13 @@ class LightGBMTrainer(GBDTTrainer):
     for features with the categorical data type, consider using the
     :class:`Categorizer` preprocessor to set the dtypes in the dataset.
 
+    .. note::
+        ``LightGBMTrainer`` does not modify or otherwise alter the working
+        of the LightGBM distributed training algorithm.
+        Ray only provides orchestration, data ingest and fault tolerance.
+        For more information on LightGBM distributed training, refer to
+        `LightGBM documentation <https://lightgbm.readthedocs.io/>`__.
+
     Example:
         .. code-block:: python
 

python/ray/train/xgboost/xgboost_trainer.py

@@ -20,6 +20,13 @@ class XGBoostTrainer(GBDTTrainer):
     This Trainer runs the XGBoost training loop in a distributed manner
     using multiple Ray Actors.
 
+    .. note::
+        ``XGBoostTrainer`` does not modify or otherwise alter the working
+        of the XGBoost distributed training algorithm.
+        Ray only provides orchestration, data ingest and fault tolerance.
+        For more information on XGBoost distributed training, refer to
+        `XGBoost documentation <https://xgboost.readthedocs.io>`__.
+
     Example:
         .. code-block:: python