From c9297a4f569950da06a5b54088e5092ed07a7977 Mon Sep 17 00:00:00 2001 From: Haibin Lin Date: Mon, 26 Nov 2018 15:52:26 -0800 Subject: [PATCH] Update row_sparse.md --- docs/tutorials/sparse/row_sparse.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/tutorials/sparse/row_sparse.md b/docs/tutorials/sparse/row_sparse.md index 27cc0d3d903e..46a5edad075e 100644 --- a/docs/tutorials/sparse/row_sparse.md +++ b/docs/tutorials/sparse/row_sparse.md @@ -459,7 +459,7 @@ Note that warning messages will be printed when such a storage fallback event ha ## Sparse Optimizers -In MXNet, sparse gradient updates are applied when weight, state and gradient are all in `row_sparse` storage. +In MXNet, sparse gradient updates are applied when gradient is in `row_sparse` storage and the optimizer is created with `lazy_update=True`. The sparse optimizers only update the row slices of the weight and the states whose indices appear in `gradient.indices`. For example, the default update rule for SGD optimizer is: