Torch7: A Matlab-like Environment for Machine Learning

By Ronan Collobert, Koray Kavukcuoglu, Clément Farabet. 2011

1,610 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

Convolutional Sequence to Sequence Learning (2017, methodology): “All models are implemented in Torch (Collobert et al., 2011) and trained on a single Nvidia M40 GPU except for WMT’14 English-French for which we use a multi-GPU setup on a single machine.”
Caffe: Convolutional Architecture for Fast Feature Embedding (2014, background): “…discontinued Decaf [2] BSD Python discontinued OverFeat [9] unspeciﬁed Lua C++,Python centralized Theano/Pylearn2 [4] BSD Python distributed Torch7 [1] BSD Lua distributed While Caﬀe was ﬁrst designed for vision, it has been adopted and improved by users in speech recognition, robotics,…”
Pylearn2: a machine learning research library (2013, methodology): “For example, in Torch [9] or DistBelief [14] the Model s are modular, but to train a layer, the layer needs to implement at the very least a backpropogation method for computing derivatives.”
Theano: new features and speed improvements (2012, methodology): “Section 4 compares Theano’s performance with that of Torch7 (Collobert et al., 2011) on neural network benchmarks, and RNNLM (Mikolov et al., 2011) on recurrent neural network benchmarks.”
Character-level Convolutional Networks for Text Classification (2015, methodology): “The implementation is done using Torch 7 [4].”
Wide Residual Networks (2016, neutral): “Our implementation is based on Torch [6].”
PyTorch: An Imperative Style, High-Performance Deep Learning Library (2019, background): “Specialized libraries such as cuDNN [22], along with a body of academic work (such as [23] and [24]), produced a set of high-performance reusable deep learning kernels that enabled frameworks such as Caffe [1], Torch7 [25], or TensorFlow [3] to take advantage of these hardware accelerators .”
Predicting Deep Zero-Shot Convolutional Neural Networks Using Textual Descriptions (2015, methodology): “Our model implementation is based on the open-source package Torch [1].”
Deep Networks with Stochastic Depth (2016, positive): “Our implementations are in Torch 7 [31].”
Recurrent Instance Segmentation (2015, methodology): “We perform two kinds of experiments, in which we study the capabilities of our approach to both segment and count instances.”
Learning to decompose for object detection and instance segmentation (2015, methodology): “All implementations were based on torch deep learning library(Collobert et al., 2011).”
Speed/Accuracy Trade-Offs for Modern Convolutional Object Detectors (2016, background): “, DistBelief [7], Caffe [18], Torch [4]) and different hardware.”
A Convolutional Encoder Model for Neural Machine Translation (2016, methodology): “Gradients of convolutional layers are scaled by sqrt(dim(input))−1 similar to Collobert et al. (2011b). We use dropout on the embeddings and decoder outputs hi with a rate of 0.”
Quantized Neural Networks: Training Neural Networks with Low Precision Weights and Activations (2016, methodology): “We thank the developers of Torch, (Collobert et al., 2011) a Lua based environment, and Theano (Bergstra et al.”
Prioritized Experience Replay (2015, methodology): “The complete architecture is shown in Figure 6, and is implemented using Torch7 (Collobert et al., 2011).”
Densely Connected Convolutional Networks (2016, neutral): “rate the effectiveness of DenseNet on several benchmark datasets and compare it with state-of-the-art network architectures, especially with ResNet and its variants. The implementation is in Torch 7 (Collobert et al., 2011). The code to reproduce the results is available at https://github.com/liuzhu”
Perceptual Losses for Real-Time Style Transfer and Super-Resolution (2016, methodology): “Our implementation uses Torch [52] and cuDNN [53]; training takes roughly 4 hours on a single GTX Titan X GPU.”
Language Modeling with Gated Convolutional Networks (2016, methodology): “We implement our models in Torch (Collobert et al., 2011) and train on Tesla M40 GPUs.”