address comments: improve API description

Author: XilunWu

prototype_source/context_parallel.rst

@@ -33,15 +33,16 @@ Two Ring Attention variants have been implemented: `the all-gather based pass-KV
     local query tensor chunk. Our modified all-gather based pass-KV algorithm concurrently all-gathers KV shards and computes attention output for the local query tensor chunk
     using local key and value tensor chunks, followed by a final computation of attention output for the local query tensor and remaining KV shards. This allows some degree of
     overlap between the attention computation and the all-gather collective.
-2.  The all-to-all approach uses interleaved all-to-all collectives to ring shuffle KV shards to overlap the SDPA computation and the all-to-all communication
+2.  The all-to-all approach uses interleaved all-to-all collectives to ring shuffle KV shards to overlap the SDPA (Scaled Dot Product Attention) computation and the all-to-all communication
     necessary for the next SDPA.
 
 The Context Parallel APIs consist of two parts:
 
-1. ``context_parallel()`` allows users to create a Python context where the SDPA function (``torch.nn.functional.scaled_dot_product_attention``)
-   will be automatically replaced with Ring Attention. To shard Tensors along a dimension, simply pass the Tensors and their sharding dimensions to
-   argument ``buffers`` and ``buffer_seq_dims`` respectively.
-2. ``set_rotate_method()`` allows users to choose between the all-gather based pass-KV approach and the all-to-all based pass-KV approach.
+1.  ``context_parallel()`` allows users to create a Python context where the SDPA function (``torch.nn.functional.scaled_dot_product_attention``)
+    will be automatically replaced with Ring Attention. To shard Tensors along a dimension, simply pass the Tensors and their sharding dimensions to
+    argument ``buffers`` and ``buffer_seq_dims`` respectively. We recommend that users add tensors computing along the sequence dimension to ``buffers``
+    and shard them along this dimension.
+2.  ``set_rotate_method()`` allows users to choose between the all-gather based pass-KV approach and the all-to-all based pass-KV approach.
 
 
 Setup