go>> udacity_ml_engineer_nano>> 返回
项目作者： d88w

项目描述 ：
[SageMaker] [xgboost] training and deploying ML models on Amazon SageMaker
高级语言： Jupyter Notebook
项目地址: git://github.com/d88w/udacity_ml_engineer_nano.git
创建时间： 2020-06-17T05:20:28Z
项目社区：https://github.com/d88w/udacity_ml_engineer_nano
开源协议：
下载


udacity_ml_engineer_nano

FOLDER - 0 Software Fundamentals

  • Some basic items on optimizing code and OOP syntax
  • Creating Python packages

FOLDER - 1 Machine Learning in SageMaker

  • Deployment of ML models in Amazon SageMaker
  • 1.01_Boston Housing - XGBoost - High Level
    • Building a model using Batch Transform in SageMaker - use the Python SDK to interact with SageMaker
  • 1.02_IMDB Sentiment Analysis - XGBoost
    • Building a model using Batch Transform in SageMaker - use the Python SDK to interact with SageMaker
  • 1.03_Boston Housing - XGBoost - Low Level
    • Building a model using SageMaker
    • Low level approach where we describe different tasks we want SageMaker to perform
    • The high level approach makes developing new models very straightforward, requiring very little code. The reason this can be done is that certain decisions have been made for you.
    • The low level approach allows you to be far more particular in how you want the various tasks executed, which is good for when you want to do something a little more complicated.
  • 1.04_Boston Housing - XGBoost (Deploy) - High Level
    • Building and Deploying simple model using the Python SDK to interact with SageMaker
  • 1.05_Boston Housing - XGBoost (Deploy) - Low Level
    • Building and Deploying simple model with the low level approach
    • Using the low level approach to deploy our model requires us to create an endpoint, which will be used to send data to our model and to get inference results.
    • In order to create an endpoint in SageMaker, we first need to describe an endpoint configuration.
  • 1.06_IMDB Sentiment Analysis - XGBoost - Web App
    • Deploy the sentiment model via a web app that
    • Using Amazon Lambda and API Gateway to solve: a) the authentication issue, b) bag of words encoding expected by model vs the block of text in the web app
    • Flow is as following: App <<—>> API & Lamda <<—>> Model
    • App finished product: https://derekspublicbucket.s3.us-east-2.amazonaws.com/index_sentiment.html
    • This app only works when I turn on the model & endpoint in AWS Sagemaker