Frameworks and Libraries

Examples on how to use different frameworks on SageMaker.

Apache MXNet

• Building an image embedding server with Gluon
• Sentiment Analysis with Apache MXNet and Gluon
• Implementing a Recommender System with SageMaker, MXNet, and Gluon
• Importing and hosting an ONNX model with MXNet
• Exporting ONNX Models with MXNet
• Using the Apache MXNet Module API with SageMaker Training and Batch Transformation
• Using Amazon Elastic Inference with MXNet on Amazon SageMaker

Deep Graph Library

• Graph convolutional matrix completion hyperparameter tuning with Amazon SageMaker and Deep Graph Library with MXNet backend
• Training Amazon SageMaker models for molecular property prediction by using DGL with PyTorch backend
• Hyperparameter tuning with Amazon SageMaker for molecular property prediction
• Training Amazon SageMaker models by using the Deep Graph Library with MXNet backend
• Output
• Hyperparameter tuning with Amazon SageMaker and Deep Graph Library with MXNet backend
• Training Amazon SageMaker models by using the Deep Graph Library with PyTorch backend
• Output
• Hyperparameter tuning with Amazon SageMaker and Deep Graph Library with PyTorch backend

PyTorch

• MNIST training with PyTorch
• Deploy a Trained PyTorch Model
• Word-level language modeling using PyTorch
• Word-level language modeling using PyTorch

R

• R Examples
  • Using R with Amazon SageMaker - Basic Notebook
  • Using R in SageMaker Processing
  • Batch Transform Using R with Amazon SageMaker
  • Hyperparamter Optimization Using R with Amazon SageMaker
  • Hyperparameter Tuning Your Own R Algorithm with Your Own Container in Amazon SageMaker

Scikit-learn

• Develop, Train, Optimize and Deploy Scikit-Learn Random Forest
• Prepare data
• Writing a Script Mode script
• Local training
• SageMaker Training
• Deploy to a real-time endpoint
• Don’t forget to delete the endpoint !
• Develop, Train, Register and Batch Transform Scikit-Learn Random Forest
• Prepare data
• Writing a Script Mode script
• Launching a SageMaker training job with the Python SDK
• Launch the 1st training job
• Create a Model Package Group for the trained model to be registered
• Register the model of the 1st training job in the Model Registry
• Create a transform job with the default configurations from the model of the 1st training job
• Launch the 2nd training job
• Register the model of 2nd training job in the Model Registry
• Create a transform job with the default configurations from the model of the 2nd training job
• Conclusion

TensorFlow

• Training a Tensorflow Model on MNIST
• Deploy a Trained TensorFlow V2 Model
• Migrating scripts from Framework Mode to Script Mode
• Construct an entry point script for training
• Deploy the trained model to prepare for predictions
• Invoking the endpoint
• Clean-up
• Migrating scripts from Framework Mode to Script Mode
• Construct an entry point script for training
• Deploy the trained model to prepare for predictions
• Invoking the endpoint
• Clean-up
• Horovod Distributed Training with SageMaker TensorFlow script mode.
• TensorFlow Script Mode with Pipe Mode Input
• Running training using the Python SDK
• Using TensorFlow Scripts in SageMaker - Quickstart
• Training in SageMaker
• TensorFlow Script Mode - Using Shell scripts
• Getting the image for training
• Training in SageMaker
• TensorFlow Eager Execution with Amazon SageMaker Script Mode and Automatic Model Tuning
• Prepare dataset
• TensorFlow script mode training and serving
• Set up the environment
• Construct a script for distributed training
• Create a training job using the TensorFlow estimator
• Deploy the trained model to an endpoint
• Invoke the endpoint
• Delete the endpoint
• Visualize Amazon SageMaker Training Jobs with TensorBoard
• TensorBoard
• Cleaning up

JAX

• Training and Deploying ML Models using JAX on SageMaker