|
Amazon SageMaker JumpStart is a machine studying (ML) hub that may make it easier to speed up your ML journey. SageMaker JumpStart offers you entry to built-in algorithms with pre-trained fashions from fashionable mannequin hubs, pre-trained basis fashions that can assist you carry out duties corresponding to article summarization and picture technology, and end-to-end options to unravel frequent use circumstances.
Right this moment, I’m pleased to announce you could now share ML artifacts, corresponding to fashions and notebooks, extra simply with different customers that share your AWS account utilizing SageMaker JumpStart.
Utilizing SageMaker JumpStart to Share ML Artifacts
Machine studying is a crew sport. You would possibly need to share your fashions and notebooks with different knowledge scientists in your crew to collaborate and enhance productiveness. Or, you would possibly need to share your fashions with operations groups to place your fashions into manufacturing. Let me present you how one can share ML artifacts utilizing SageMaker JumpStart.
In SageMaker Studio, choose Fashions within the left navigation menu. Then, choose Shared fashions and Shared by my group. Now you can uncover and search ML artifacts that different customers shared inside your AWS account. Notice you could add and share ML artifacts developed with SageMaker in addition to these developed exterior of SageMaker.
To share a mannequin or pocket book, choose Add. For fashions, present primary data, corresponding to title, description, knowledge sort, ML job, framework, and any extra metadata. This data helps different customers to seek out the suitable fashions for his or her use circumstances. You can too allow coaching and deployment in your mannequin. This enables customers to fine-tune your shared mannequin and deploy the mannequin in only a few clicks by way of SageMaker JumpStart.
To allow mannequin coaching, you’ll be able to choose an current SageMaker coaching job that can autopopulate all related data. This data consists of the container framework, coaching script location, mannequin artifact location, occasion sort, default coaching and validation datasets, and goal column. You can too present customized mannequin coaching data by choosing a prebuilt SageMaker Deep Studying Container or choosing a customized Docker container in Amazon ECR. You can too specify default hyperparameters and metrics for mannequin coaching.
To allow mannequin deployment, you additionally must outline the container picture to make use of, the inference script and mannequin artifact location, and the default occasion sort. Take a look on the SageMaker Developer Information to study extra about mannequin coaching and mannequin deployment choices.
Sharing a pocket book works equally. You have to present primary details about your pocket book and the Amazon S3 location of the pocket book file.
Customers that share your AWS account can now browse and choose shared fashions to fine-tune, deploy endpoints, or run notebooks straight in SageMaker JumpStart.
In SageMaker Studio, choose Fast begin options within the left navigation menu, then choose Options, fashions, instance notebooks to entry all shared ML artifacts, along with pre-trained fashions from fashionable mannequin hubs and end-to-end options.
Now Obtainable
The brand new ML artifact-sharing functionality inside Amazon SageMaker JumpStart is out there immediately in all AWS Areas the place Amazon SageMaker JumpStart is out there. To study extra, go to Amazon SageMaker JumpStart and the SageMaker JumpStart documentation.
Begin sharing your fashions and notebooks with Amazon SageMaker JumpStart immediately!
— Antje
