Tuesday, 29 May 2018

Amazon SageMaker Algorithms supports Pipe Input mode for faster training jobs

With the built-in Amazon SageMaker algorithms, you can now operate training jobs with the built-in algorithms up to 35 percent faster with the Pipe input mode. By utilizing the pipe input mode the training jobs stream data directly from the Amazon S3 to the algorithm container on the training instance so to provider faster start time for the training jobs and also for better throughput. Most of the Amazon SageMaker algorithms work well when the optimized protobuf recordIO format is used for training data for speed optimization. 

No comments:

Post a Comment

Amazon Aurora Serverless Accessible in more Regions

Amazon Aurora Serverless is an on-demand, auto-scaling configuration for Amazon Aurora is a MySQL and PostgreSQL-compatible relational da...