Torch: a modular machine learning software library

By Ronan Collobert, Samy Bengio, J. Mariéthoz. 2002

565 citations recorded by the source corpus.

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In ML Systems & Hardware

Citations in the corpus

PyTorch: An Imperative Style, High-Performance Deep Learning Library (2019, methodology): “The autograd [16] package popularized the use of this technique for NumPy arrays, and similar approaches are used in frameworks such as Chainer [5], DyNet [7], Lush [14], Torch [6], Jax [17] and Flux.”
TensorFlow: A system for large-scale machine learning (2016, neutral): “ six-core Intel Core i7-5930K CPU at 3.5GHz and an NVIDIA Titan X GPU. Training step time (ms) Library AlexNet Overfeat OxfordNet GoogleNet Caffe [36] 324 823 1068 1935 Neon [56] 87 211 320 270 Torch [17] 81 268 529 470 TensorFlow 81 279 540 445 Table 1: Step times for training four convolutional mo”
Aggregated Residual Transformations for Deep Neural Networks (2016, methodology): “Grouped convolutions are supported by Caffe [19], Torch [3], and other libraries, mainly for compatibility of AlexNet.”
TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems (2016, neutral): “Theano [7], Torch [13], Caffe [26], Chainer [49] and the Computational Network Toolkit [54] are a few systems designed primarily for the training of neural networks.”
Towards Better Analysis of Deep Convolutional Neural Networks (2016, background): “Common deep learning frameworks include Caffe [28], Theano [6], Torch [10], and TensorFlow [1].”
The hardware lottery (2020, background): “It was only in the 2000’s that there started to be a more healthy ecosystem around software developed for deep neural network approaches with the emergence of LUSH (Lecun & Bottou, 2002) and subsequently TORCH (Collobert et al., 2002).”
Superneurons: dynamic GPU memory management for training deep neural networks (2018, methodology): “Therefore, most DL systems parallelize the training with Data Parallelism for better performance [2, 5, 9, 14].”
Between-Class Learning for Image Classification (2017, methodology): “To validate the comparison, we also incorporated BC learning into their original Torch [6] training codes 4 .”
Evolino for recurrent support vector machines (2005, methodology): “The support vector method used to compute the weights ( wij in equation 2) is a large scale approximation of the quadratic constrained optimization, as impleme nted in [35].”
Cloud-Based or On-Device: An Empirical Study of Mobile Deep Inference (2017, background): “There are a number of popular deep learning frameworks, such as Caffe [10], Torch [11], TensorFlow [12], that ease the training and deploying of deep learning models.”
DyNet: The Dynamic Neural Network Toolkit (2017, methodology): “These tools, including Theano [7], TensorFlow [1], Torch [13], CNTK [64], MxNet [10], and Chainer [62] provide neural network function primitives (e.g., linear algebra operations, non-linear transformations, etc.), parameter initialization and optimization routines, and usually the ability to…”
JAX, M.D.: End-to-End Differentiable, Hardware Accelerated, Molecular Dynamics in Pure Python (2019, methodology): “To address these issues, several projects developed adapters (Artrith & Urban, 2016; Artrith et al., 2017; Lot et al., 2019; Onat et al., 2018) between common ML languages, like Torch and Tensorflow, and common MD languages like LAMMPS.”
A Survey on Deep Learning (2018, background): “Table 2 lists a smattering of popular deep learning frameworks for architecture designs, such as Caffe [72], DeepLearning4j (DL4j) [143], Torch [25], Neon [69], Theano [5], MXNet [17], TensorFlow [1], and Microsoft Cognitive Toolkit (CNTK) [173].”
GossipGraD: Scalable Deep Learning using Gossip Communication based Asynchronous Gradient Descent (2018, methodology): “Several DL implementations such as TensorFlow [5], Caffe [6], Theano [7, 8], and Torch [9] have become available.”
Radio Machine Learning Dataset Generation with GNU Radio (2016, background): “…architectures such Keras (Chollet, 2015), Theano (Bergstra et al., 2010), TensorFlow (Abadi et al., 2015), Caffe (Jia et al., 2014), and Torch (Collobert et al., 2002) which all provide collections of learning primitives, routines and automatic gradient computation for training with novel…”
HyperDrive: exploring hyperparameters with POP scheduling (2017, background): “In recent years many machine-learning frameworks have been introduced to help developers build and train machine learning models for solving different artificial intelligence tasks across supervised, unsupervised, and reinforcement learning domains [3, 6, 9, 10, 15, 31].”