Today, we are excited to announce that the Mistral 7B foundation models, developed by Mistral AI, are available for customers through Amazon SageMaker JumpStart to deploy with one click for running inference. With 7 billion parameters, Mistral 7B can be easily customized and quickly deployed. You can try out this model with SageMaker JumpStart, a machine learning (ML) hub that provides access to algorithms and models so you can quickly get started with ML. In this post, we walk through how to discover and deploy the Mistral 7B model.

What is Mistral 7B

Mistral 7B is a foundation model developed by Mistral AI, supporting English text and code generation abilities. It supports a variety of use cases, such as text summarization, classification, text completion, and code completion. To demonstrate the easy customizability of the model, Mistral AI has also released a Mistral 7B Instruct model for chat use cases, fine-tuned using a variety of publicly available conversation datasets.

Mistral 7B is a transformer model and uses grouped-query attention and sliding-window attention to achieve faster inference (low latency) and handle longer sequences. Group query attention is an architecture that combines multi-query and multi-head attention to achieve output quality close to multi-head attention and comparable speed to multi-query attention. Sliding-window attention uses the stacked layers of a transformer to attend in the past beyond the window size to increase context length. Mistral 7B has an 8,000-token context length, demonstrates low latency and high throughput, and has strong performance when compared to larger model alternatives, providing low memory requirements at a 7B model size. The model is made available under the permissive Apache 2.0 license, for use without restrictions.

What is SageMaker JumpStart

With SageMaker JumpStart, ML practitioners can choose from a growing list of best-performing foundation models. ML practitioners can deploy foundation models to dedicated Amazon SageMaker instances within a network isolated environment, and customize models using SageMaker for model training and deployment.

You can now discover and deploy Mistral 7B with a few clicks in Amazon SageMaker Studio or programmatically through the SageMaker Python SDK, enabling you to derive model performance and MLOps controls with SageMaker features such as Amazon SageMaker Pipelines, Amazon SageMaker Debugger, or container logs. The model is deployed in an AWS secure environment and under your VPC controls, helping ensure data security.

Discover models

You can access Mistral 7B foundation models through SageMaker JumpStart in the SageMaker Studio UI and the SageMaker Python SDK. In this section, we go over how to discover the models in SageMaker Studio.

SageMaker Studio is an integrated development environment (IDE) that provides a single web-based visual interface where you can access purpose-built tools to perform all ML development steps, from preparing data to building, training, and deploying your ML models. For more details on how to get started and set up SageMaker Studio, refer to Amazon SageMaker Studio.

In SageMaker Studio, you can access SageMaker JumpStart, which contains pre-trained models, notebooks, and prebuilt solutions, under Prebuilt and automated solutions.

From the SageMaker JumpStart landing page, you can browse for solutions, models, notebooks, and other resources. You can find Mistral 7B in the Foundation Models: Text Generation carousel.

You can also find other model variants by choosing Explore all Text Models or searching for “Mistral.”

You can choose the model card to view details about the model such as license, data used to train, and how to use. You will also find two buttons, Deploy and Open notebook, which will help you use the model (the following screenshot shows the Deploy option).

Deploy models

Deployment starts when you choose Deploy. Alternatively, you can deploy through the example notebook that shows up when you choose Open notebook. The example notebook provides end-to-end guidance on how to deploy the model for inference and clean up resources.

To deploy using notebook, we start by selecting the Mistral 7B model, specified by the model_id. You can deploy any of the selected models on SageMaker with the following code:

from sagemaker.jumpstart.model import JumpStartModel

model = JumpStartModel(model_id="huggingface-llm-mistral-7b-instruct")
predictor = model.deploy()

This deploys the model on SageMaker with default configurations, including default instance type (ml.g5.2xlarge) and default VPC configurations. You can change these configurations by specifying non-default values in JumpStartModel. After it’s deployed, you can run inference against the deployed endpoint through the SageMaker predictor:

payload = {"inputs": "<s>[INST] Hello! [/INST]"}
predictor.predict(payload)

Optimizing the deployment configuration

Mistral models use Text Generation Inference (TGI version 1.1) model serving. When deploying models with the TGI deep learning container (DLC), you can configure a variety of launcher arguments via environment variables when deploying your endpoint. To support the 8,000-token context length of Mistral 7B models, SageMaker JumpStart has configured some of these parameters by default: we set MAX_INPUT_LENGTH and MAX_TOTAL_TOKENS to 8191 and 8192, respectively. You can view the full list by inspecting your model object:

print(model.env)

By default, SageMaker JumpStart doesn’t clamp concurrent users via the environment variable MAX_CONCURRENT_REQUESTS smaller than the TGI default value of 128. The reason is because some users may have typical workloads with small payload context lengths and want high concurrency. Note that the SageMaker TGI DLC supports multiple concurrent users through rolling batch. When deploying your endpoint for your application, you might consider whether you should clamp MAX_TOTAL_TOKENS or MAX_CONCURRENT_REQUESTS prior to deployment to provide the best performance for your workload:

model.env["MAX_CONCURRENT_REQUESTS"] = "4"

Here, we show how model performance might differ for your typical endpoint workload. In the following tables, you can observe that small-sized queries (128 input words and 128 output tokens) are quite performant under a large number of concurrent users, reaching token throughput on the order of 1,000 tokens per second. However, as the number of input words increases to 512 input words, the endpoint saturates its batching capacity—the number of concurrent requests allowed to be processed simultaneously—resulting in a throughput plateau and significant latency degradations starting around 16 concurrent users. Finally, when querying the endpoint with large input contexts (for example, 6,400 words) simultaneously by multiple concurrent users, this throughput plateau occurs relatively quickly, to the point where your SageMaker account will start encountering 60-second response timeout limits for your overloaded requests.

Inference and example prompts

Mistral 7B

You can interact with a base Mistral 7B model like any standard text generation model, where the model processes an input sequence and outputs predicted next words in the sequence. The following is a simple example with multi-shot learning, where the model is provided with several examples and the final example response is generated with contextual knowledge of these previous examples:

> Input
Tweet: "I get sad when my phone battery dies."
Sentiment: Negative
###
Tweet: "My day has been :+1:"
Sentiment: Positive
###
Tweet: "This is the link to the article"
Sentiment: Neutral
###
Tweet: "This new music video was incredibile"
Sentiment:

> Output
 Positive

Mistral 7B instruct

The instruction-tuned version of Mistral accepts formatted instructions where conversation roles must start with a user prompt and alternate between user and assistant. A simple user prompt may look like the following:

<s>[INST] {user_prompt} [/INST]

A multi-turn prompt would look like the following:

<s>[INST] {user_prompt_1} [/INST] {assistant_response_1} </s><s>[INST] {user_prompt_1} [/INST]

This pattern repeats for however many turns are in the conversation.

In the following sections, we explore some examples using the Mistral 7B Instruct model.

Knowledge retrieval

The following is an example of knowledge retrieval:

> Input
<s>[INST] Which country has the most natural lakes? Answer with only the country name. [/INST] 

> Output
1. Canada

Large context question answering

To demonstrate how to use this model to support large input context lengths, the following example embeds a passage, titled “Rats” by Robert Sullivan (reference), from the MCAS Grade 10 English Language Arts Reading Comprehension test into the input prompt instruction and asks the model a directed question about the text:

> Input
<s>[INST] A rat is a rodent, the most common mammal in the world. Rattus norvegicus is one of the approximately four hundred different kinds of rodents, and it is known by many names, each of which describes a trait or a perceived trait or sometimes a habitat: the earth rat, the roving rat, the barn rat, the fi eld rat, the migratory rat, the house rat, the sewer rat, the water rat, the wharf rat, the alley rat, the gray rat, the brown rat, and the common rat. The average brown rat is large and stocky; it grows to be approximately sixteen inches long from its nose to its tail—the size of a large adult human male’s foot—and weighs about a pound, though brown rats have been measured by scientists and exterminators at twenty inches and up to two pounds. The brown rat is sometimes confused with the black rat, or Rattus rattus, which is smaller and once inhabited New York City and all of the cities of America but, since Rattus norvegicus pushed it out, is now relegated to a minor role. (The two species still survive alongside each other in some Southern coastal cities and on the West Coast, in places like Los Angeles, for example, where the black rat lives in attics and palm trees.) The black rat is always a very dark gray, almost black, and the brown rat is gray or brown, with a belly that can be light gray, yellow, or even a pure-seeming white. One spring, beneath the Brooklyn Bridge, I saw a red-haired brown rat that had been run over by a car. Both pet rats and laboratory rats are Rattus norvegicus, but they are not wild and therefore, I would emphasize, not the subject of this book. Sometimes pet rats are called fancy rats. But if anyone has picked up this book to learn about fancy rats, then they should put this book down right away; none of the rats mentioned herein are at all fancy.

Rats are nocturnal, and out in the night the brown rat’s eyes are small and black and shiny; when a fl ashlight shines into them in the dark, the eyes of a rat light up like the eyes of a deer. Though it forages* in darkness, the brown rat has poor eyesight. It makes up for this with, fi rst of all, an excellent sense of smell. . . . They have an excellent sense of taste, detecting the most minute amounts of poison, down to one part per million. A brown rat has strong feet, the two front paws each equipped with four clawlike nails, the rear paws even longer and stronger. It can run and climb with squirrel-like agility. It is an excellent swimmer, surviving in rivers and bays, in sewer streams and toilet bowls.

The brown rat’s teeth are yellow, the front two incisors being especially long and sharp, like buckteeth. When the brown rat bites, its front two teeth spread apart. When it gnaws, a fl ap of skin plugs the space behind its incisors. Hence, when the rat gnaws on indigestible materials—concrete or steel, for example—the shavings don’t go down the rat’s throat and kill it. Its incisors grow at a rate of fi ve inches per year. Rats always gnaw, and no one is certain why—there are few modern rat studies. It is sometimes erroneously stated that the rat gnaws solely to limit the length of its incisors, which would otherwise grow out of its head, but this is not the case: the incisors wear down naturally. In terms of hardness, the brown rat’s teeth are stronger than aluminum, copper, lead, and iron. They are comparable to steel. With the alligator-like structure of their jaws, rats can exert a biting pressure of up to seven thousand pounds per square inch. Rats, like mice, seem to be attracted to wires—to utility wires, computer wires, wires in vehicles, in addition to gas and water pipes. One rat expert theorizes that wires may be attractive to rats because of their resemblance to vines and the stalks of plants; cables are the vines of the city. By one estimate, 26 percent of all electric-cable breaks and 18 percent of all phone-cable disruptions are caused by rats. According to one study, as many as 25 percent of all fi res of unknown origin are rat-caused. Rats chew electrical cables. Sitting in a nest of tattered rags and newspapers, in the fl oorboards of an old tenement, a rat gnaws the head of a match—the lightning in the city forest.

When it is not gnawing or feeding on trash, the brown rat digs. Anywhere there is dirt in a city, brown rats are likely to be digging—in parks, in fl owerbeds, in little dirt-poor backyards. They dig holes to enter buildings and to make nests. Rat nests can be in the floorboards of apartments, in the waste-stuffed corners of subway stations, in sewers, or beneath old furniture in basements. “Cluttered and unkempt alleyways in cities provide ideal rat habitat, especially those alleyways associated with food-serving establishments,” writes Robert Corrigan in Rodent Control, a pest control manual. “Alley rats can forage safely within the shadows created by the alleyway, as well as quickly retreat to the safety of cover in these narrow channels.” Often, rats burrow under concrete sidewalk slabs. Entrance to a typical under-the-sidewalk rat’s nest is gained through a two-inch-wide hole—their skeletons collapse and they can squeeze into a hole as small as three quarters of an inch wide, the average width of their skull. This tunnel then travels about a foot down to where it widens into a nest or den. The den is lined with soft debris, often shredded plastic garbage or shopping bags, but sometimes even grasses or plants; some rat nests have been found stuffed with the gnawed shavings of the wood-based, spring-loaded snap traps that are used in attempts to kill them. The back of the den then narrows into a long tunnel that opens up on another hole back on the street. This second hole is called a bolt hole; it is an emergency exit. A bolt hole is typically covered lightly with dirt or trash—camoufl age. Sometimes there are networks of burrows, which can stretch beneath a few concrete squares on a sidewalk, or a number of backyards, or even an entire city block—when Rattus norvegicus fi rst came to Selkirk, England, in 1776, there were so many burrows that people feared the town might sink. Rats can also nest in basements, sewers, manholes, abandoned pipes of any kind, fl oorboards, or any hole or depression. “Often,” Robert Corrigan writes, “‘city rats’ will live unbeknownst to people right beneath their feet.”

Rats also inhabit subways, as most people in New York City and any city with a subway system are well aware. Every once in a while, there are reports of rats boarding trains, but for the most part rats stay on the tracks—subway workers I have talked to refer to rats as “track rabbits.” People tend to think that the subways are fi lled with rats, but in fact rats are not everywhere in the system; they live in the subways according to the supply of discarded human food and sewer leaks. Sometimes, rats use the subway purely for nesting purposes; they fi nd ways through the walls of the subway stations leading from the tracks to the restaurants and stores on the street—the vibrations of subway trains tend to create rat-size cracks and holes. Many subway rats tend to live near stations that are themselves near fast-food restaurants. At the various subway stations near Herald Square, for example, people come down from the streets and throw the food that they have not eaten onto the tracks, along with newspapers and soda bottles and, I have noticed, thousands of nolonger-charged AA batteries, waiting to leak acid. The rats eat freely from the waste and sit at the side of the little streams of creamy brown sewery water that fl ows between the rails. They sip the water the way rats do, either with their front paws or by scooping it up with their incisors.


What is the most likely reason the author states, "cables are the vines of the city"? [/INST] 


> Output
1. The author states that rats are attracted to wires, such as utility wires, computer wires, and vehicle wires, which may be due to their resemblance to vines and the stalks of plants.

Mathematics and reasoning

The Mistral models also report strengths in mathematics accuracy. Mistral can provide comprehension such as the following math logic:

> Input
<s>[INST] I bought an ice cream for 6 kids. Each cone was $1.25 and I paid with a $10 bill. How many dollars did I get back? Explain first before answering. [/INST]

> Output
1. First, let's figure out how much the ice cream cost in total. Since each cone was $1.25 and there were 6 kids, the total cost would be:

Total cost = Number of cones * Cost per cone
Total cost = 6 * $1.25
Total cost = $7.50

2. Next, we'll subtract the total cost from the amount paid ($10) to find out how much money was returned:

Money returned = Amount paid - Total cost
Money returned = $10 - $7.50
Money returned = $2.50

So, you got back $2.50 after buying the ice cream for the 6 kids.

Coding

The following is an example of a coding prompt:

> Input
<s>[INST] In Bash, how do I list all text files in the current directory (excluding subdirectories) that have been modified in the last month? [/INST] 


> Output
1. Use the `find` command to search for text files in the current directory (excluding subdirectories) that have been modified in the last month. The `-type f` option specifies that we only want to search for files, not directories. The `-name "*.txt"` option specifies that we only want to search for files with the `.txt` extension. The `-mtime -30` option specifies that we only want to search for files that have been modified in the last 30 days (1 month).

```bash
find . -type f -name "*.txt" -mtime -30
```

This command will list all text files in the current directory (excluding subdirectories) that have been modified in the last month.

Clean up

After you’re done running the notebook, make sure to delete all the resources that you created in the process so your billing is stopped. Use the following code:

predictor.delete_model()
predictor.delete_endpoint()

Conclusion

In this post, we showed you how to get started with Mistral 7B in SageMaker Studio and deploy the model for inference. Because foundation models are pre-trained, they can help lower training and infrastructure costs and enable customization for your use case. Visit Amazon SageMaker JumpStart now to get started.

Resources

• SageMaker JumpStart documentation
• SageMaker JumpStart foundation models documentation
• SageMaker JumpStart product detail page
• SageMaker JumpStart model catalog

About the Authors

Dr. Kyle Ulrich is an Applied Scientist with the Amazon SageMaker JumpStart team. His research interests include scalable machine learning algorithms, computer vision, time series, Bayesian non-parametrics, and Gaussian processes. His PhD is from Duke University and he has published papers in NeurIPS, Cell, and Neuron.

Dr. Ashish Khetan is a Senior Applied Scientist with Amazon SageMaker JumpStart and helps develop machine learning algorithms. He got his PhD from University of Illinois Urbana-Champaign. He is an active researcher in machine learning and statistical inference, and has published many papers in NeurIPS, ICML, ICLR, JMLR, ACL, and EMNLP conferences.

Vivek Singh is a product manager with Amazon SageMaker JumpStart. He focuses on enabling customers to onboard SageMaker JumpStart to simplify and accelerate their ML journey to build generative AI applications.

Roy Allela is a Senior AI/ML Specialist Solutions Architect at AWS based in Munich, Germany. Roy helps AWS customers—from small startups to large enterprises—train and deploy large language models efficiently on AWS. Roy is passionate about computational optimization problems and improving the performance of AI workloads.

Read More

According to Gartner, 85% of software buyers trust online reviews as much as personal recommendations. Customers provide feedback and reviews about products they have purchased through many channels, including review websites, vendor websites, sales calls, social media, and many others. The problem with the increasing volume of customer reviews across multiple channels is that it can be challenging for companies to process and derive meaningful insights from the data using traditional methods. Machine learning (ML) can analyze large volumes of product reviews and identify patterns, sentiments, and topics discussed. With this information, companies can gain a better understanding of customer preferences, pain points, and satisfaction levels. They can also use this information to improve products and services, identify trends, and take strategic actions that drive business growth. However, implementing ML can be a challenge for companies that lack resources such as ML practitioners, data scientists, or artificial intelligence (AI) developers. With the new Amazon SageMaker Canvas features, business analysts can now use ML to derive insights from product reviews.

SageMaker Canvas is designed for the functional needs of business analysts to use AWS no code ML for ad hoc analysis of tabular data. SageMaker Canvas is a visual, point-and-click service that allows business analysts to generate accurate ML predictions without writing a single line of code or requiring ML expertise. You can use models to make predictions interactively and for batch scoring on bulk datasets. SageMaker Canvas offers fully-managed ready-to-use AI model and custom model solutions. For common ML use cases, you can use a ready-to-use AI model to generate predictions with your data without any model training. For ML use cases specific to your business domain, you can train an ML model with your own data for custom prediction.

In this post, we demonstrate how to use the ready-to-use sentiment analysis model and custom text analysis model to derive insights from product reviews. In this use case, we have a set of synthesized product reviews that we want to analyze for sentiments and categorize the reviews by product type, to make it easy to draw patterns and trends that can help business stakeholders make better informed decisions. First, we describe the steps to determine the sentiment of the reviews using the ready-to-use sentiment analysis model. Then, we walk you through the process to train a text analysis model to categorize the reviews by product type. Next, we explain how to review the trained model for performance. Finally, we explain how to use the trained model to perform predictions.

Sentiment analysis is a natural language processing (NLP) ready-to-use model that analyzes text for sentiments. Sentiment analysis may be run for single line or batch predictions. The predicted sentiments for each line of text are either positive, negative, mixed or neutral.

Text analysis allows you to classify text into two or more categories using custom models. In this post, we want to classify product reviews based on product type. To train a text analysis custom model, you simply provide a dataset consisting of the text and the associated categories in a CSV file. The dataset requires a minimum of two categories and 125 rows of text per category. After the model is trained, you can review the model’s performance and retrain the model if needed, before using it for predictions.

Prerequisites

Complete the following prerequisites:

1. Have an AWS account.
2. Set up SageMaker Canvas.
3. Download the sample product reviews datasets:
   • sample_product_reviews.csv – Contains 2,000 synthesized product reviews and is used for sentiment analysis and Text Analysis predictions.
   • sample_product_reviews_training.csv – Contains 600 synthesized product reviews and three product categories, and is for text analysis model training.

Sentiment analysis

First, you use sentiment analysis to determine the sentiments of the product reviews by completing the following steps.

1. On the SageMaker console, click Canvas in the navigation pane, then click Open Canvas to open the SageMaker Canvas application.
2. Click Ready-to-use models in the navigation pane, then click Sentiment analysis.
3. Click Batch prediction, then click Create dataset.
4. Provide a Dataset name and click Create.
5. Click Select files from your computer to import the sample_product_reviews.csv dataset.
6. Click Create dataset and review the data. The first column contains the reviews and is used for sentiment analysis. The second column contains the review ID and is used for reference only.
7. Click Create dataset to complete the data upload process.
8. In the Select dataset for predictions view, select sample_product_reviews.csv and then click Generate predictions. 
9. When the batch prediction is complete, click View to view the predictions.

The Sentiment and Confidence columns provide the sentiment and confidence score, respectively. A confidence score is a statistical value between 0 and 100%, that shows the probability that the sentiment is correctly predicted.

10. Click Download CSV to download the results to your computer.

Text analysis

In this section, we go through the steps to perform text analysis with a custom model: importing the data, training the model and then making predictions.

Import the data

First import the training dataset. Complete the following steps:

1. On Ready-to-use models page, click Create a custom model
2. For Model name, enter a name (for example, Product Reviews Analysis). Click Text analysis, then click Create.
3. On the Select tab, click Create dataset to import the sample_product_reviews_training.csv dataset.
4. Provide a Dataset name and click Create.
5. Click Create dataset and review the data. The training dataset contains a third column describing product category, the target column consisting of three products: books, video, and music.
6. Click Create dataset to complete the data upload process.
7. On the Select dataset page, select sample_product_reviews_training.csv and click Select dataset.

Train the model

Next, you configure the model to begin the training process.

1. On the Build tab, on the Target column drop-down menu, click product_category as the training target.
2. Click product_review as the source.
3. Click Quick build to start the model training.

For more information about the differences between Quick build and Standard build, refer to Build a custom model.

When the model training is complete, you may review the performance of the model before you use it for prediction.

4. On the Analyze tab, the model’s confidence score will be displayed. A confidence score indicates how certain a model is that its predictions are correct. On the Overview tab, review the performance for each category.
5. Click Scoring to review the model accuracy insights.
6. Click Advance metrics to review the confusion matrix and F1 score.

Make predictions

To make a prediction with your custom model, complete the following steps:

1. On the Predict tab, click Batch prediction, then click Manual.
2. Click the same dataset, sample_product_reviews.csv, that you used previously for the sentiment analysis, then click Generate predictions.
3. When the batch prediction is complete, click View to view the predictions.

For custom model prediction, it takes some time for SageMaker Canvas to deploy the model for initial use. SageMaker Canvas automatically de-provisions the model if idle for 15 minutes to save costs.

The Prediction (Category) and Confidence columns provide the predicted product categories and confidence scores, respectively.

4. Highlight the completed job, select the three dots and click Download to download the results to your computer.

Clean up

Click Log out in the navigation pane to log out of the SageMaker Canvas application to stop the consumption of Canvas session hours and release all resources.

Conclusion

In this post, we demonstrated how you can use Amazon SageMaker Canvas to derive insights from product reviews without ML expertise. First, you used a ready-to-use sentiment analysis model to determine the sentiments of the product reviews. Next, you used text analysis to train a custom model with the quick build process. Finally, you used the trained model to categorize the product reviews into product categories. All without writing a single line of code. We recommend that you repeat the text analysis process with the standard build process to compare the model results and prediction confidence.

About the Authors

Gavin Satur is a Principal Solutions Architect at Amazon Web Services. He works with enterprise customers to build strategic, well-architected solutions and is passionate about automation. Outside work, he enjoys family time, tennis, cooking and traveling.

Les Chan is a Sr. Solutions Architect at Amazon Web Services, based in Irvine, California. Les is passionate about working with enterprise customers on adopting and implementing technology solutions with the sole focus of driving customer business outcomes. His expertise spans application architecture, DevOps, serverless, and machine learning.

Aaqib Bickiya is a Solutions Architect at Amazon Web Services based in Southern California. He helps enterprise customers in the retail space accelerate projects and implement new technologies. Aaqib’s focus areas include machine learning, serverless, analytics, and communication services

Read More

A recommendation engine is only as good as the data used to prepare it. Transforming raw data into a format that is suitable for a model is key to getting better personalized recommendations for end-users.

In this post, we walk through how to prepare and import the MovieLens dataset, a dataset prepared by GroupLens research at the University of Minnesota, which consists of a variety of user rankings of various movies, into Amazon Personalize using Amazon SageMaker Data Wrangler. [1]

Solution overview

Amazon Personalize is a managed service whose core value proposition is its ability to learn user preferences from their past behavior and quickly adjust those learned preferences to take account of changing user behavior in near-real time. To be able to develop this understanding of users, Amazon Personalize needs to train on the historical user behavior so that it can find patterns that are generalizable towards the future. Specifically, the main type of data that Amazon Personalize learns from is what we call an interactions dataset, which is a tabular dataset that consists of at minimum three critical columns, userID, itemID, and timestamp, representing a positive interaction between a user and an item at a specific time. Users Amazon Personalize need to upload data containing their own customer’s interactions in order for the model to be able to learn these behavioral trends. Although the internal algorithms within Amazon Personalize have been chosen based on Amazon’s experience in the machine learning space, a personalized model doesn’t come pre-loaded with any sort of data and trains models on a customer-by-customer basis.

The MovieLens dataset explored in this walkthrough isn’t in this format, so to prepare it for Amazon Personalize, we use SageMaker Data Wrangler, a purpose-built data aggregation and preparation tool for machine learning. It has over 300 preconfigured data transformations as well as the ability to bring in custom code to create custom transformations in PySpark, SQL, and a variety of data processing libraries, such as pandas.

Prerequisites

First, we need to have an Amazon SageMaker Studio domain set up. For details on how to set it up, refer to Onboard to Amazon SageMaker Domain using Quick setup.

Also, we need to set up the right permissions using AWS Identity and Access Management (IAM) for Amazon Personalize and Amazon SageMaker service roles so that they can access the needed functionalities.

You can create a new Amazon Personalize dataset group to use in this walkthrough or use an existing one.

Finally, we need to download and unzip the MovieLens dataset and place it in an Amazon Simple Storage Service (Amazon S3) bucket.

Launch SageMaker Data Wrangler from Amazon Personalize

To start with the SageMaker Data Wrangler integration with Amazon Personalize, complete the following steps:

1. On the Amazon Personalize console, navigate to the Overview page of your dataset group.
2. Choose Import interaction data, Import user data, or Import item data, depending on the dataset type (for this post, we choose Import interaction data).

3. For Import method, select Import data using Data Wrangler.
4. Choose Next.

5. Specify the SageMaker domain, user profile, and IAM service role that you created earlier as part of the prerequisites.
6. Choose Next.

7. Continue through the steps to launch an instance of SageMaker Data Wrangler.

Setting up the environment for the first time can take up to 5 minutes.

Import the raw data into SageMaker Data Wrangler

When using SageMaker Data Wrangler to prepare and import data, we use a data flow. A data flow defines a series of transformations and analyses on data to prepare it to create a machine learning model. Each time we add a step to our flow, SageMaker Data Wrangler takes an action on our data, such as joining it with another dataset or dropping some rows and columns.

To start, let’s import the raw data.

1. On the data flow page, choose Import data.

With SageMaker Data Wrangler, we can import data from over 50 supported data sources.

2. For Data sources¸ choose Amazon S3.

3. Choose the dataset you uploaded to your S3 bucket.

SageMaker Data Wrangler automatically displays a preview of the data.

4. Keep the default settings and choose Import.

After the data is imported, SageMaker Data Wrangler automatically validates the datasets and detects the data types for all the columns based on its sampling.

5. Choose Data flow at the top of the Data types page to view the main data flow before moving to the next step.

One of the main advantages of SageMaker Data Wrangler is the ability to run previews of your transformations on a small subset of data before committing to apply the transformations on the entire dataset. To run the same transformation on multiple partitioned files in Amazon S3, you can use parameters.

Transform the data

To transform data in SageMaker Data Wrangler, you add a Transform step to your data flow. SageMaker Data Wrangler includes over 300 transforms that you can use to prepare your data, including a Map columns for Amazon Personalize transform. You can use the general SageMaker Data Wrangler transforms to fix issues such as outliers, type issues, and missing values, or apply data preprocessing steps.

To use Amazon Personalize, the data you provided in the interactions dataset must match your dataset schema. For our movie recommender engine, the proposed interactions dataset schema includes:

• user_id (string)
• item_id (string)
• event_type (string)
• timestamp (in Unix epoch time format)

To learn more about Amazon Personalize datasets and schemas, refer to Datasets and schemas.

The ratings.csv file as shown in the last step in the previous section includes movies rated from 1–5. We want to build a movie recommender engine based on that. To do so, we must complete the following steps:

1. Modify the columns data types.
2. Create two event types: Click and Watch.
3. Assign all movies rated 2 and above as Click and movies rated 4 and above as both Click and Watch.
4. Drop the ratings column.
5. Map the columns to the Amazon Personalize interactions dataset schema.
6. Validate that our timestamp is in Unix epoch time.

Note that Step 3 isn’t needed to make a personalization model. If we want to use one of the Amazon Personalize streamlined video on demand domain recommenders, such as Top Picks for You, Click and Watch would be required event types. However, if we don’t have these, we could not include an event type field (or add our own event types such as the raw user ratings) and use a custom recipe such as User Personalization. Regardless of what type of recommendation engine we use, we need to ensure our dataset contains only representations of positive user intent. So whatever approach you choose, you need to drop all one-star ratings (and possibly two-star ratings as well).

Now let’s use SageMaker Data Wrangler to perform the preceding steps.

1. On the Data flow page, choose the first transform, called Data types.

2. Update the type for each column.
3. Choose Preview to reflect the changes, then choose Update.

4. To add a step in the data flow, choose the plus sign next to the step you want to perform the transform on, then choose Add transform.

5. To filter the event Click out of the movie ratings, we add a Filter data step to filter out the movies rated 2 and above.

6. Add another custom transform step to add a new column, eventType, with Click as an assigned value.
7. Choose Preview to review your transformation to double-check the results are as intended, then choose Add.
8. In this case, we write some PySpark code to add a column called eventType whose value will be uniformly Click for all of our two-star through five-star movies:

from pyspark.sql.functions import lit

df = df.withColumn("eventType", lit("Click"))

9. For the Watch events, repeat the previous steps for movies rated 4 and above and assign the Watch value by adding the steps to the Data types step. Our PySpark code for these steps is as follows:

from pyspark.sql.functions import lit

df = df.withColumn("eventType", lit("Watch"))

Up to this point, the data flow should look like the following screenshot.

Concatenate datasets

Because we have two datasets for watched and clicked events, let’s walk through how to concatenate these into one interactions dataset.

1. On the Data flow page, choose the plus sign next to Create Watch Event and choose Concatenate.

2. Choose the other final step (Create Click Event), and this should automatically map (converge) both the sets into a concatenate preview.

3. Choose Configure to view a preview of the concatenated datasets.
4. Add a name to the step.
5. Choose Add to add the step.

The data flow now looks like the following screenshot.

6. Now, let’s add a Manage columns step to drop the original rating column.

Amazon Personalize has default column names for users, items, and timestamps. These default column names are user_id, item_id, and timestamp.

7. Let’s add a Transform for Amazon Personalize step to replace the existing column headers with the default headers.
8. In our case, we also use the event_type field, so let’s map that as well.

With this step, the data transformation activity is complete and the interactions dataset is ready for the next step.

Next, let’s validate our timestamps.

9. We can do this by adding a Custom transform step. For this post, we choose Python (User-Defined Function).
10. Choose timestamp as the input column and as the output, create a new column called readable_timestamp.
11. Choose Python as the mode for the transformation and insert the following code for the Python function:

def custom_func(value: int) → str:
    return datetime.utcfromtimestamp(value).strftime('%Y-%m-%d %H:%M:%S')

12. Choose Preview to review the changes.

In this case, we see dates in the 2000s—because MovieLens started collecting data in 1996, this aligns with what is expected. If we don’t choose Add, this transformation won’t be added to our data flow.

13. Because this was merely a sanity check, you can navigate back to the data flow by choosing Data flow in the upper left corner.

Finally, we add an analysis step to create a summary report about the dataset. This step performs an analysis to assess the suitability of the dataset for Amazon Personalize.

14. Choose the plus sign next to the final step on the data flow and choose Add analysis.
15. For Analysis type¸ choose Data Quality And Insights Report for Amazon Personalize.
16. For Dataset type¸ choose Interactions.
17. Choose Create.

The MovieLens dataset is quite clean, so the analysis shows no issues. If some issues were identified, you can iterate on the dataset and rerun the analysis until you can address them.

Note the analysis by default runs on a sample of 50,000 rows.

Import the dataset to Amazon Personalize

At this point, our raw data has been transformed and we are ready to import the transformed interactions dataset to Amazon Personalize. SageMaker Data Wrangler gives you the ability to export your data to a location within an S3 bucket. You can specify the location using one of the following methods:

• Destination node – Where SageMaker Data Wrangler stores the data after it has processed it
• Export to – Exports the data resulting from a transformation to Amazon S3
• Export data – For small datasets, you can quickly export the data that you’ve transformed

With the Destination node method, to export your data, you create destination nodes and a SageMaker Data Wrangler job. Creating a SageMaker Data Wrangler job starts a SageMaker Processing job to export your flow. You can choose the destination nodes that you want to export after you’ve created them.

1. Choose the plus sign next to the node that represents the transformations you want to export.

2. Choose Export to and then choose Amazon S3 (via Jupyter Notebook).

Note we could have also chosen to export the data to Amazon Personalize via a Jupyter notebook available in SageMaker Data Wrangler.

3. For Dataset name, enter a name, which will be used as a folder name in the S3 bucket provided as a destination.
4. You can specify the file type, field delimiter, and compression method.
5. Optionally, specify the number of partitions and column to partition by.
6. Choose Add destination.

The data flow should look like the following screenshot.

7. Create a job to process the data flow and store the data in the destination (S3 bucket) that we configured in the previous step.
8. Enter a job name, then choose Configure job.

SageMaker Data Wrangler provides the ability to configure the instance type, instance count, and job configuration, and the ability to create a schedule to process the job. For guidance on how to choose an instance count, refer to Create and Use a Data Wrangler Flow.

To monitor the status of the job, navigate to the Dashboard page on the SageMaker console. The Processing section shows the number of completed and created jobs. You can drill down to get more details about the completed job.

When the job is complete, a new file of the transformed data is created in the destination specified.

9. Return to the Amazon Personalize console and navigate to the dataset group to import another dataset.
10. Choose Import interaction data.

11. Select Import data directly into Amazon Personalize datasets to import the transformed dataset directly from Amazon S3, then choose Next.

12. Define the schema. For this post, our case our dataset consists of the user_id (string), item_id (string), event_type (string), and timestamp (long) fields.

At this point, you can create a video on demand domain recommender or a custom solution. To do so, follow the steps in Preparing and importing data

Conclusion

In this post, we described how to use SageMaker Data Wrangler to prepare a sample dataset for Amazon Personalize. SageMaker Data Wrangler offers over 300 transformations. These transformations and the ability to add custom user transformations can help streamline the process of creating a quality dataset to offer hyper-personalized content to end-users.

Although we only explored how to prepare an interactions dataset in this post, you can use SageMaker Data Wrangler to prepare user and item datasets as well. For more information on the types of data that can be used with Amazon Personalize, refer to Datasets and schemas.

If you’re new to Amazon Personalize or SageMaker Data Wrangler, refer to Get Started with Amazon Personalize or Get Started with SageMaker Data Wrangler, respectively. If you have any questions related to this post, please add them in the comments section.

About the Authors

Maysara Hamdan is a Partner Solutions Architect based in Atlanta, Georgia. Maysara has over 15 years of experience in building and architecting Software Applications and IoT Connected Products in Telecom and Automotive Industries. In AWS, Maysara helps partners in building their cloud practices and growing their businesses. Maysara is passionate about new technologies and is always looking for ways to help partners innovate and grow.

Eric Bolme is a Specialist Solution Architect with AWS based on the East Coast of the United States. He has 8 years of experience building out a variety of deep learning and other AI use cases and focuses on Personalization and Recommendation use cases with AWS.

References

Read More