DeBertaV3

<pre>A tutorial for implementing the DeBERTaV3 model On AWS<br><br>1. Introduction<br>This project implements a Machine Learning Pipeline with DeBERTaV3.<br><br>2. Code files<br>Here, we introduce main files for the ML pipeline of DeBERTaV3.<br>(a) sagemaker-pa-mlops-modelbuild/codebuild-buildspec.yml: it builds the ML continue-integrate pipeline with scripts in the sagemaker-pa-mlops-modelbuild folder.<br>(b) sagemaker-pa-mlops-modelbuild/pipelines: it builds nd register the ML pipeline.<br>(c) sagemaker-pa-mlops-modelbuild/pipelines/run_pipeline.py: it builds nd register the ML pipeline.<br>(d) sagemaker-pa-mlops-modelbuild/pipelines/DeBERTaV3/train.py: it trains the DeBERTaV3 model.<br>(e) sagemaker-pa-mlops-modelbuild/pipelines/DeBERTaV3/evaluate.py: it evaluates the DeBERTaV3 model to report accuracy and f1-score.<br><br>3. Dataset and preprocess files<br>We provided test datasets for running the proposed MCGCN model, which contains one-day traffic speed, spatial context, and temporal context datasets in the folder “./raw_data/speed_origin_data/”. The detail of each file and its formats are as follows:<br>(a) Data-preprocess/CityofYarra_Verification_Data.xlsx: City of Yarra Dataset.<br>(b) Data-preprocess/Whitehorse_classification.xlsx: City of Whitehorse Dataset.<br>(c) Data-preprocess/high priority.csv: Planning applications with high priority Dataset.<br>(a) Data-preprocess/Split.ipynb: It splits the dataset.</pre>

基于亚马逊云服务（AWS）实现DeBERTaV3模型的教程

1. 简介
本项目基于DeBERTaV3搭建机器学习流水线（Machine Learning Pipeline）。

2. 代码文件
本节将介绍DeBERTaV3机器学习流水线的核心文件。
(a) sagemaker-pa-mlops-modelbuild/codebuild-buildspec.yml：该文件用于基于sagemaker-pa-mlops-modelbuild文件夹内的脚本构建机器学习持续集成（Continuous Integration）流水线。
(b) sagemaker-pa-mlops-modelbuild/pipelines：该目录用于构建并注册机器学习流水线。
(c) sagemaker-pa-mlops-modelbuild/pipelines/run_pipeline.py：该脚本用于构建并注册机器学习流水线。
(d) sagemaker-pa-mlops-modelbuild/pipelines/DeBERTaV3/train.py：该脚本用于训练DeBERTaV3模型。
(e) sagemaker-pa-mlops-modelbuild/pipelines/DeBERTaV3/evaluate.py：该脚本用于对DeBERTaV3模型进行评估，以输出准确率与F1值。

3. 数据集与预处理文件
我们提供了用于运行所提出的MCGCN模型的测试数据集，相关数据集包含单日交通速度、空间上下文及时域上下文数据，存储于“./raw_data/speed_origin_data/”文件夹中。各文件的详细信息与格式如下：
(a) Data-preprocess/CityofYarra_Verification_Data.xlsx：雅拉市数据集。
(b) Data-preprocess/Whitehorse_classification.xlsx：怀特霍斯市数据集。
(c) Data-preprocess/high priority.csv：高优先级规划申请数据集。
(d) Data-preprocess/Split.ipynb：该脚本用于对数据集进行划分。