arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1802.04730cs.PLcs.LG
收藏

Tensor Comprehensions:与框架无关的高性能机器学习抽象

Tensor Comprehensions: Framework-Agnostic High-Performance Machine Learning Abstractions

Facebook AI Research(Facebook人工智能研究院);Inria(法国国家信息与自动化研究所);ENS(巴黎高等师范学院);ETH Zürich(苏黎世联邦理工学院);MIT CSAIL(麻省理工学院计算机科学与人工智能实验室);Polly Labs(Polly实验室)

Nicolas Vasilache, Oleksandr Zinenko, Theodoros Theodoridis, Priya Goyal, Zachary DeVito, William S. Moses, Sven Verdoolaege, Andrew Adams, Albert Cohen

摘要:

具有卷积和递归网络的深度学习模型现在无处不在,可以分析大量的音频、图像、视频、文本和图形数据,应用于自动翻译、语音到文本、场景理解、用户偏好排名、广告投放等。构建这些网络的竞争框架,如TensorFlow、Chainer、CNTK、Torch/PyTorch、Caffe 1/2、MXNet和Theano,探索可用性和表现力、面向研究或生产以及支持的硬件之间的不同权衡。它们在计算运算符的DAG上运行,包装高性能库,如CUDNN(用于NVIDIA GPU)或NNPACK(用于各种CPU),并自动分配内存,同步,分发。如果计算不适合现有的高性能库调用,则需要自定义运算符,这通常需要很高的工程成本。当研究人员发明新的算子时,这是经常需要的:这样的算子遭受严重的性能惩罚,这限制了创新的步伐。此外,即使这些框架可以使用现有的运行时调用,它通常也不能为用户的特定网络架构和数据集提供最佳性能,缺少运算符之间的优化以及可以在知道数据大小和形状的情况下进行的优化。我们的贡献包括(1)一种接近深度学习数学的语言,称为张量理解(Tensor Comprehension),(2)一个多面体即时编译器,用于将深度学习DAG的数学描述转换为具有委托存储管理和同步的CUDA内核,还提供了优化,如针对特定大小的运算符融合和专业化,(3)由自动调谐器填充的编译缓存。[摘要截止]

英文摘要:

Deep learning models with convolutional and recurrent networks are now ubiquitous and analyze massive amounts of audio, image, video, text and graph data, with applications in automatic translation, speech-to-text, scene understanding, ranking user preferences, ad placement, etc. Competing frameworks for building these networks such as TensorFlow, Chainer, CNTK, Torch/PyTorch, Caffe1/2, MXNet and Theano, explore different tradeoffs between usability and expressiveness, research or production orientation and supported hardware. They operate on a DAG of computational operators, wrapping high-performance libraries such as CUDNN (for NVIDIA GPUs) or NNPACK (for various CPUs), and automate memory allocation, synchronization, distribution. Custom operators are needed where the computation does not fit existing high-performance library calls, usually at a high engineering cost. This is frequently required when new operators are invented by researchers: such operators suffer a severe performance penalty, which limits the pace of innovation. Furthermore, even if there is an existing runtime call these frameworks can use, it often doesn't offer optimal performance for a user's particular network architecture and dataset, missing optimizations between operators as well as optimizations that can be done knowing the size and shape of data. Our contributions include (1) a language close to the mathematics of deep learning called Tensor Comprehensions, (2) a polyhedral Just-In-Time compiler to convert a mathematical description of a deep learning DAG into a CUDA kernel with delegated memory management and synchronization, also providing optimizations such as operator fusion and specialization for specific sizes, (3) a compilation cache populated by an autotuner. [Abstract cutoff]

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑