🐾SageMaker Training cost optimisation🐾

Five ways to optimise SageMaker training cost:

1️⃣  Use GPU in the right cases: it is better with simple calculations that can be done in parallel. CPU is better for single complex calculations. Consider using Amazon Elastic Inference to attach low-cost GPU-powered acceleration to Sagemaker instances and reduce the cost of Deep Learning training by up to 75%.

2️⃣  Monitor your resources utilization using Amazon Cloud Watch and change instance type appropriately if it is underutilized. SageMaker Debugger also can improve your ML model's compute resources utilization.

3️⃣  Setup automatic shut down of idle resources. Instructions for setup and scripts can be found in GitHub repository.

4️⃣  You can save up to 90% using Managed Spot Training.

5️⃣  Use Savings Plans, they offer up to 72% savings over the on-demand price, in exchange for your commitment to use a specific amount of compute power.

---

If you like this post, you can share APAWS newsletter with friends: