aboutsummaryrefslogtreecommitdiffhomepage
path: root/tensorflow/contrib/distribute/README.md
diff options
context:
space:
mode:
Diffstat (limited to 'tensorflow/contrib/distribute/README.md')
-rw-r--r--tensorflow/contrib/distribute/README.md3
1 files changed, 2 insertions, 1 deletions
diff --git a/tensorflow/contrib/distribute/README.md b/tensorflow/contrib/distribute/README.md
index 91a27f97b7..2e025765e4 100644
--- a/tensorflow/contrib/distribute/README.md
+++ b/tensorflow/contrib/distribute/README.md
@@ -231,7 +231,8 @@ The same `input_fn` will be used for all workers if you use
important to shuffle your dataset in your `input_fn`.
`MirroredStrategy` will insert a `tf.dataset.Dataset.shard` call in you
-`input_fn`. As a result, each worker gets a fraction of your input data.
+`input_fn` if `auto_shard_dataset` is set to `True`. As a result, each worker
+gets a fraction of your input data.
### Performance Tips