astha/languagemodelsforRNNdecomposition

---
license: mit
---

This repository is for the paper "Decomposing a Recurrent Neural Network into Modules for Enabling Reusability and Replacement". To use the data, there are two directories:

1. language datasets: Contains the necessary Tatoeba files used for the experiments. We have experimented with 4 languages(English, French, Italian and German).
2. language_models: Contains all trained language models and scripts to train them. It's organized in this way: language_models/{X}: contains language models for X, X={Vanilla RNN,LSTM,GRU}.

language_models/{X}/motivating example models/problem1: Contains models needed to reproduce motivating example Problem 1.
language_models/{X}/motivating example models/problem2: Contains models needed to reproduce motivating example Problem 2.
language_models/{X}/reuse models: Contains combined models needed for reuse experiments.
language_models/{X}/rq1 models: Contains all models needed for rq1.
language_models/{X}/training script: Contains the training script to get the models.

Note: To run the training scripts, direct the source path to /language datasets/Tatoeba/<name.txt> depending on the requirement.

原始信息汇总

数据集概述

目录结构

1. language datasets: 包含用于实验的Tatoeba文件，实验涉及四种语言（英语、法语、意大利语和德语）。
2. language_models: 包含所有训练好的语言模型和训练脚本，具体组织如下：
   • language_models/{X}: 包含针对X的语言模型，X={Vanilla RNN, LSTM, GRU}。
     • motivating example models/problem1: 包含用于重现动机示例问题1的模型。
     • motivating example models/problem2: 包含用于重现动机示例问题2的模型。
     • reuse models: 包含用于重用实验的组合模型。
     • rq1 models: 包含用于rq1的所有模型。
     • training script: 包含获取模型的训练脚本。

使用说明

• 运行训练脚本时，需将源路径指向/language datasets/Tatoeba/<name.txt>，具体取决于需求。