diff --git a/README.md b/README.md
index 051f5ab..1f44239 100644
--- a/README.md
+++ b/README.md
@@ -1,9 +1,10 @@
 
 # Introduction
-**The implementations of quantization and de-quantization operators are now available in [Caffe2/Pytorch 1.0](https://github.com/pytorch/pytorch/blob/master/caffe2/operators/fused_rowwise_random_quantization_ops.cc) ([an example](https://github.com/pytorch/pytorch/blob/master/caffe2/python/operator_test/rand_quantization_op_test.py)). TernGrad is landed into Facebook AI platforms in production to overcome communication bottleneck in large-scale training.**
 
 This repo is the TensorFlow code for our oral paper in NIPS 2017 ([TernGrad: Ternary Gradients to Reduce Communication in Distributed Deep Learning](https://papers.nips.cc/paper/6749-terngrad-ternary-gradients-to-reduce-communication-in-distributed-deep-learning.pdf)).
 
+**The implementations of quantization and de-quantization operators are now available in [Caffe2/Pytorch 1.0](https://github.com/pytorch/pytorch/blob/master/caffe2/operators/fused_rowwise_random_quantization_ops.cc) ([an example](https://github.com/pytorch/pytorch/blob/master/caffe2/python/operator_test/rand_quantization_op_test.py)). TernGrad is landed into Facebook AI platforms in production to overcome communication bottleneck in large-scale training.**
+
 [video](https://www.youtube.com/watch?v=WWWQXTb_69c&feature=youtu.be&t=20s), [poster](/Poster_Wen_NIPS2017.pdf), [slides](/NIPS17-TernGrad-slides-v3.pdf)
 
 This is a modified copy of TensorFlow [inception](https://github.com/tensorflow/models/tree/master/inception) (with original contributions kept).