From 29072a5f8efa1193be4dc9ced74690b1cd43a5a0 Mon Sep 17 00:00:00 2001
From: lnstadrum <21985366+lnstadrum@users.noreply.github.com>
Date: Wed, 8 Nov 2023 17:37:36 +0100
Subject: [PATCH] Update README.md

---
 README.md | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/README.md b/README.md
index 0114b04..ae401ea 100644
--- a/README.md
+++ b/README.md
@@ -1,6 +1,6 @@
 # Overview
 
-`PixelShuffle`/`PixelUnshuffle` in PyTorch and `depth_to_space`/`space_to_depth` in TensorFlow are very similar, but they are not numerically identical unless the upsampled image contains a single channel. This can be easyly verified by hand: while the tensor dimensions match between the PyTorch and TensorFlow worlds, the output channels do not follow the same order.
+`PixelShuffle`/`PixelUnshuffle` in PyTorch and `depth_to_space`/`space_to_depth` in TensorFlow are very similar, but they are not numerically identical unless the upsampled image contains a single channel. This can be easily verified by hand: while the tensor dimensions match between the PyTorch and TensorFlow worlds, the output channels do not follow the same order.
 
 However, in some deployment setups, there might be performance benefits to using the space-to-depth/depth-to-space variant. For example, at the moment of writing, [Android NN API](https://developer.android.com/ndk/guides/neuralnetworks) only supports depth-to-space and space-to-depth.
 
@@ -29,4 +29,4 @@ from s2d2s import SpaceToDepth, DepthToSpace
 
 module = SpaceToDepth(2)
 y = module(x)
-```
\ No newline at end of file
+```