-
Optimizing Transformers with Approximate Computing for Faster, Smaller and more Accurate NLP Models
Transformer models have garnered a lot of interest in recent years by de...
read it
-
EvalNorm: Estimating Batch Normalization Statistics for Evaluation
Batch normalization (BN) has been very effective for deep learning and i...
read it
-
Double Forward Propagation for Memorized Batch Normalization
Batch Normalization (BN) has been a standard component in designing deep...
read it
-
Towards Stabilizing Batch Statistics in Backward Propagation of Batch Normalization
Batch Normalization (BN) is one of the most widely used techniques in De...
read it
-
A New Look at Ghost Normalization
Batch normalization (BatchNorm) is an effective yet poorly understood te...
read it
-
HyperEmbed: Tradeoffs Between Resources and Performance in NLP Tasks with Hyperdimensional Computing enabled Embedding of n-gram Statistics
Recent advances in Deep Learning have led to a significant performance i...
read it
-
On Layer Normalization in the Transformer Architecture
The Transformer is widely used in natural language processing tasks. To ...
read it
Rethinking Batch Normalization in Transformers
The standard normalization method for neural network (NN) models used in Natural Language Processing (NLP) is layer normalization (LN). This is different than batch normalization (BN), which is widely-adopted in Computer Vision. The preferred use of LN in NLP is principally due to the empirical observation that a (naive/vanilla) use of BN leads to significant performance degradation for NLP tasks; however, a thorough understanding of the underlying reasons for this is not always evident. In this paper, we perform a systematic study of NLP transformer models to understand why BN has a poor performance, as compared to LN. We find that the statistics of NLP data across the batch dimension exhibit large fluctuations throughout training. This results in instability, if BN is naively implemented. To address this, we propose Power Normalization (PN), a novel normalization scheme that resolves this issue by (i) relaxing zero-mean normalization in BN, (ii) incorporating a running quadratic mean instead of per batch statistics to stabilize fluctuations, and (iii) using an approximate backpropagation for incorporating the running statistics in the forward pass. We show theoretically, under mild assumptions, that PN leads to a smaller Lipschitz constant for the loss, compared with BN. Furthermore, we prove that the approximate backpropagation scheme leads to bounded gradients. We extensively test PN for transformers on a range of NLP tasks, and we show that it significantly outperforms both LN and BN. In particular, PN outperforms LN by 0.4/0.6 BLEU on IWSLT14/WMT14 and 5.6/3.0 PPL on PTB/WikiText-103.
READ FULL TEXT
Comments
There are no comments yet.