Update tutorial about placing sharding_filter (pytorch#487)

Summary: See the feedback from a user: pytorch#454 (comment) We should explicitly ask users to place `sharding_filter` as early as possible. Pull Request resolved: pytorch#487 Reviewed By: wenleix Differential Revision: D36812259 Pulled By: ejguan fbshipit-source-id: 4c983f3216a80be398f85b20871e65b0e41627e0
ejguan · Jun 2, 2022 · b20fbf0 · b20fbf0
1 parent 74c5ba3
commit b20fbf0
Showing 1 changed file with 5 additions and 0 deletions.
diff --git a/docs/source/tutorial.rst b/docs/source/tutorial.rst
@@ -176,6 +176,11 @@ When we re-run, we will get:
     ...
     n_sample = 6
 
+Note:
+
+- Place ``ShardingFilter`` (``datapipe.sharding_filter``) as early as possible in the pipeline, especially before expensive
+  operations such as decoding, in order to avoid repeating these expensive operations across worker/distributed processes.
+
 
 You can find more DataPipe implementation examples for various research domains `on this page <torchexamples.html>`_.