Winning the Lottery with Continuous Sparsification

Authors: Pedro Savarese, Hugo Silva, Michael Maire

NeurIPS 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical results show that we surpass the state-of-the-art for both objectives, across models and datasets, including VGG trained on CIFAR-10 and Res Net-50 trained on Image Net.
Researcher Affiliation Academia Pedro Savarese TTI-Chicago savarese@ttic.edu Hugo Silva University of Alberta hugoluis@ualberta.ca Michael Maire University of Chicago mmaire@uchicago.edu
Pseudocode Yes Algorithm 1 Iterative Magnitude Pruning [19] Input: Pruning ratio τ, number of rounds R, iterations per round T, rewind point k... Algorithm 2 Continuous Sparsification Input: Mask init s(0), penalty λ, number of rounds R, iterations per round T, rewind point k
Open Source Code Yes Code available at https://github.com/lolemacs/continuous-sparsification
Open Datasets Yes VGG trained on CIFAR-10 [23] and Res Net-50 trained on Image Net [24].
Dataset Splits No The paper describes training and testing procedures but does not explicitly mention or detail the use of a validation set for model tuning or selection.
Hardware Specification No The paper mentions using '4 GPUs' for experiments but does not specify the model or type of these GPUs, or any other specific hardware details.
Software Dependencies No The paper mentions using SGD as an optimizer but does not specify any software libraries (e.g., PyTorch, TensorFlow) or their version numbers that were used for implementation.
Experiment Setup Yes in each round, we train with SGD, a learning rate of 0.1, and a momentum of 0.9, for a total of 85 epochs, using a batch size of 64 for VGG and 128 for Res Net. We decay the learning rate by a factor of 10 at epochs 56 and 71, and utilize a weight decay of 0.0001.