Examples and Quickstarts¶
This topic contains several examples and quickstarts for common use cases for model logging and model inference in Snowflake ML. You can use these examples as a starting point for your own use case.
Beginner Quickstart¶
Getting started with Snowflake ML: train an xgboost regression model, log to model registry, and run inference in a Warehouse.
xgboost model, CPU inference in Snowpark Container Services¶
This code illustrates the key steps in deploying an XGBoost model in Snowpark Container Services (SPCS), then using the deployed model for inference.
Log a pipeline with custom preprocessing and model training¶
This example illustrates how to:
Perform feature engineering
Train a pipeline with custom preprocessing steps and an xgboost forecasting model
Run hyperparameter optimization
Log the optimum pipeline
Run inference in a arehouse or in Snowpark Container Services (SPCS)
Large scale open source embeddings model, GPU inference¶
This example uses Snowflake Notebooks on Container Runtime to train a large-scale embeddings model from the Hugging Face
sentence_transformer library and run large scale predictions using GPUs on Snowpark Container Services (SPCS).
Complete pipeline with distributed PyTorch recommender model, GPU inference¶
This example shows how to build an end-to-end distributed Pytorch recommender model using GPUs, deploying the model for GPU inference on Snowpark Container Services (SPCS).
Bring an existing model trained externally (eg. AWS Sagemaker/Azure ML/GCP Vertex AI) to Snowflake¶
These examples show how to bring your existing model in AWS Sagemaker, Azure ML, or GCP Vertex AI to Snowflake (see blog post for more details).
AWS and Azure ML Quickstart
GCP Vertex AI Quickstart
Bring an MLFlow PyFunc model to Snowflake¶
This example shows how to log an MLFlow PyFunc model in the Snowflake Model Registry and run inference.
Log a partitioned forecasting model for training and inference¶
This example shows how to log a forecasting model for running partitioned training and inference in Snowflake.
Log many-models as a collection for running partitioned inference at scale¶
This example shows how to log thousands of models as a custom partitioned model for running distributed, partitioned inference.