In this episode, we’ll dive into SageMaker AutoPilot, an AutoML capability. Starting from a tabular dataset, we’ll launch an AutoML job in just a few clicks (or just a few lines of code).
Then, we’ll explore in detail the different steps in AutoPilot, such as automatic feature engineering and model tuning. We’ll show you the auto-generated notebooks, and how you can run them yourself for further optimization.
Finally, we look at AutoGluon, an open source library for AutoML.
In the first section, we have a chat with our special guest Greg Coquillo, a Technology Risk Manager working for Amazon. We walk through an automation project that he’s currently working on for a B2B customer operating in chemicals. In order to build material safety data sheets, the project automatically extracts image and text data from over 100,000 documents a month, using AI services like Amazon Textract, Amazon Comprehend and Amazon Translate. We’ll discuss the key phases of the project, its benefits, and best practices learned along the way.
In the second section, we dive into a large-scale computer vision…
In this video, I show you how to use Savings Plans for Amazon SageMaker, a new cost optimization capability that helps you save up to 64% on your SageMaker workloads!
Companion blog post:
In this episode, we use state of the art models for natural language processing available in the Hugging Face collection. Then, we fine-tune BERT on a sentiment analysis dataset, and predict with the model. Finally, we show you how to scale your training jobs with data parallelism and model parallelism.
In this video, I show you how to fine-tune an Hugging Face model on Amazon SageMaker, and how to predict with the model on your local machine.
In this episode, we build a sentiment analysis model starting from the Amazon Customer Reviews dataset. First, we import the dataset in Parquet format in Amazon Athena. Then, we import it from Athena to SageMaker Data Wrangler for a quick look. Then, we move to a Jupyter notebook and we start engineering features using popular open source libraries (nltk and spaCy), and we automate them with SageMaker Processing. Next, we load the processed dataset in SageMaker Feature Store, both offline and online. Next, we run Athena queries on the offline store in order to build a training set, which we…
In this episode, we start from the popular Titanic survivor dataset. We import it in SageMaker Data Wrangler, where we build visualizations and apply built-in transforms (column operations, imputing missing values, one hot encoding, normalization). Then, we export these transforms to a Jupyter notebook running a SageMaker Processing job. We run the notebook and take a look at the processed dataset, before training a model with XGBoost. We also take a quick look at other export options (Python code, SageMaker Pipelines, SageMaker Feature Store). As usual, 100% live, no slides :)
Join us for future episodes at Join us for more episodes at https://amazonsagemakerfridays.splashthat.com/
In this 90-minute special, we start with a quick introduction to SageMaker, and then we walk you through the 9 SageMaker launches from AWS re:Invent 2020.
In this video, you’ll learn how deploy a web-based application for virtual proctoring, implementing face recognition and object detection with Amazon Rekognition. Very fun demo!
The code is available at https://github.com/aws-samples/amazon-rekognition-virtual-proctor
It’s that time of the year again! AWS re:Invent 2020 is over, and as it brought us quite a few new services and features, it was time for me to update my map thoroughly.
Enjoy, and please let me know if I forgot anything.