In this article, we describe an automatic differentiation module of PyTorch — a library designed to enable rapid research on machine learning models. It builds upon a few projects, most notably Lua Torch, Chainer, and HIPS Autograd [4], and provides a high performance environment with easy access to automatic differentiation of models executed on different devices (CPU and GPU). To make prototyping easier, PyTorch does not follow the symbolic approach used in many other deep learning frameworks, but focuses on differentiation of purely imperative programs, with a focus on extensibility and low overhead. Note that this preprint is a draft of certain sections from an upcoming paper covering all PyTorch features.

/pdf/automatic-differentiation-in-pytorch-gbuafkf7qo.pdf

Automatic differentiation in PyTorch

Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it provides an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. 
In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. 
We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several common benchmarks.

PyTorch: An Imperative Style, High-Performance Deep Learning Library

Probability Distributions.- Linear Models for Regression.- Linear Models for Classification.- Neural Networks.- Kernel Methods.- Sparse Kernel Machines.- Graphical Models.- Mixture Models and EM.- Approximate Inference.- Sampling Methods.- Continuous Latent Variables.- Sequential Data.- Combining Models.

Pattern Recognition and Machine Learning

Recent work has demonstrated substantial gains on many NLP tasks and benchmarks by pre-training on a large corpus of text followed by fine-tuning on a specific task. While typically task-agnostic in architecture, this method still requires task-specific fine-tuning datasets of thousands or tens of thousands of examples. By contrast, humans can generally perform a new language task from only a few examples or from simple instructions - something which current NLP systems still largely struggle to do. Here we show that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches. Specifically, we train GPT-3, an autoregressive language model with 175 billion parameters, 10x more than any previous non-sparse language model, and test its performance in the few-shot setting. For all tasks, GPT-3 is applied without any gradient updates or fine-tuning, with tasks and few-shot demonstrations specified purely via text interaction with the model. GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic. At the same time, we also identify some datasets where GPT-3's few-shot learning still struggles, as well as some datasets where GPT-3 faces methodological issues related to training on large web corpora. Finally, we find that GPT-3 can generate samples of news articles which human evaluators have difficulty distinguishing from articles written by humans. We discuss broader societal impacts of this finding and of GPT-3 in general.

/pdf/language-models-are-few-shot-learners-2fg8gvia7m.pdf

Language Models are Few-Shot Learners

Deep learning frameworks have often focused on either usability or speed, but not both. PyTorch is a machine learning library that shows that these two goals are in fact compatible: it was designed from first principles to support an imperative and Pythonic programming style that supports code as a model, makes debugging easy and is consistent with other popular scientific computing libraries, while remaining efficient and supporting hardware accelerators such as GPUs. In this paper, we detail the principles that drove the implementation of PyTorch and how they are reflected in its architecture. We emphasize that every aspect of PyTorch is a regular Python program under the full control of its user. We also explain how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance. We demonstrate the efficiency of individual subsystems, as well as the overall speed of PyTorch on several commonly used benchmarks.

/pdf/pytorch-an-imperative-style-high-performance-deep-learning-59vlz5u10y.pdf

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing

Parsing web text is progressively becoming important for many applications in natural language processing, such as machine translation, information retrieval, and sentiment analysis. Current syntactic parsing has been focused on canonical data such as newswires. When evaluated on standard benchmarks such as Wall Street Journal data set, current state-of-the-art parsers achieve accuracies well above 90%. However the accuracy drops dramatically when they are applied to new domains such as web data, barely over 80%. In order to make progress in many applications that rely on parsing, we need robust parsers that can handle such texts. One approach that is becoming popular recently is to use unsupervised word representations as extra features. Koo et al. [1] has shown that unsupervised clustering features are effective to improve dependency parsing. Turian et al. [2] examined clustering and unsupervised word embedding features on chunking and named entity recognition tasks. Unsupervised word embeddings are dense, low-dimensional and real-value vectors representing words, often induced by neural language models. They have shown that these word representation features lead to improvement in the performances. These word representations are induced by unsupervised methods, thus they are good for new domains such as the web, which has enormous amount of unlabeled data but little labeled data. In this paper we investigate the effect of unsupervised word representation features on dependency parsing with web texts. We consider two different kinds of word representations, namely Brown clustering and word embeddings induced from a neural language model. To the best of our knowledge, this is the first work that systematically examines these word representations on the task of dependency parsing on web text.

An Empirical Investigation of Word Representations for Parsing the Web

We explore best practices for training small, memory efficient machine translation models with sequence-level knowledge distillation in the domain adaptation setting. While both domain adaptation and knowledge distillation are widely-used, their interaction remains little understood. Our large-scale empirical results in machine translation (on three language pairs with three domains each) suggest distilling twice for best performance: once using general-domain data and again using in-domain data with an adapted teacher.

Distill, Adapt, Distill: Training Small, In-Domain Models for Neural Machine Translation

Cross-lingual open information extraction is the task of distilling facts from the source language into representations in the target language. We propose a novel encoder-decoder model for this problem. It employs a novel selective decoding mechanism, which explicitly models the sequence labeling process as well as the sequence generation process on the decoder side. Compared to a standard encoder-decoder model, selective decoding significantly increases the performance on a Chinese-English cross-lingual open IE dataset by 3.87-4.49 BLEU and 1.91-5.92 F1. We also extend our approach to low-resource scenarios, and gain promising improvement.

Selective Decoding for Cross-lingual Open Information Extraction

Fast inference speed is an important goal towards real-world deployment of speech translation (ST) systems. End-to-end (E2E) models based on the encoder-decoder architecture are more suitable for this goal than traditional cascaded systems, but their effectiveness regarding decoding speed has not been explored so far. Inspired by recent progress in non-autoregressive (NAR) methods in text-based translation, which generates target tokens in parallel by eliminating conditional dependencies, we study the problem of NAR decoding for E2E-ST. We propose a novel NAR E2E-ST framework, Orthros, in which both NAR and autoregressive (AR) decoders are jointly trained on the shared speech encoder. The latter is used for selecting better translation among various length candidates generated from the former, which dramatically improves the effectiveness of a large length beam with negligible overhead. We further investigate effective length prediction methods from speech inputs and the impact of vocabulary sizes. Experiments on four benchmarks show the effectiveness of the proposed method in improving inference speed while maintaining competitive translation quality compared to state-of-the-art AR E2E-ST systems.

Kevin Duh

Papers

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing

An Empirical Investigation of Word Representations for Parsing the Web

Distill, Adapt, Distill: Training Small, In-Domain Models for Neural Machine Translation

Selective Decoding for Cross-lingual Open Information Extraction

Orthros: Non-autoregressive End-to-end Speech Translation with Dual-decoder