Light Multi-Segment Activation for Model Compression

Authors: Zhenhui Xu, Guolin Ke, Jia Zhang, Jiang Bian, Tie-Yan Liu6542-6549

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on state-of-the-art NN architectures over the real-world tasks demonstrate the effectiveness and extensibility of the LMA.
Researcher Affiliation Collaboration Zhenhui Xu,1 Guolin Ke,2 Jia Zhang,2 Jiang Bian,2 Tie-Yan Liu2 1Peking University, 2Microsoft Research
Pseudocode No No pseudocode or clearly labeled algorithm blocks were found in the paper.
Open Source Code Yes 2We released the code at: https://github.com/motefly/LMA
Open Datasets Yes we first evaluate our method on CIFAR-10 and CIFAR-100, both of which are wellknown image classification datasets.
Dataset Splits Yes Open NMT integration test dataset (Ope) consisting of 200K train sentences and 10K test sentences and WMT13 (Koehn 2005) dataset for a German-English translation task.
Hardware Specification No The paper mentions running experiments on GPU ('especially running on GPU'), but does not provide specific details such as GPU models, CPU types, or memory specifications.
Software Dependencies No The paper refers to using 'Open NMT' and 'Moses project' but does not specify any software libraries or dependencies with version numbers required for reproducibility.
Experiment Setup Yes For fair comparisons, we set all the common parameters, including learning rate, batch size, hyper-parameters in distillation loss, etc., the same for all the baselines.