关于Amazon SageMaker（支持TensorFlow）是否免费的技术咨询

Hey there, let's break down the free tier details for SageMaker since you're looking to use it for TensorFlow training when your local laptop can't keep up. Both the things you heard are actually correct—SageMaker has two types of free offerings that cover those scenarios:

• 12-month limited free tier (for new AWS users)
  When you create a new AWS account, you get access to a time-bound free tier for SageMaker that lasts 12 months from sign-up. For TensorFlow specifically, this includes:

  • 250 hours per month of t2.medium notebook instance usage (great for writing, testing, and debugging your TensorFlow code)
  • 50 hours per month of ml.t2.medium training instance time (for running your actual model training jobs)
    This benefit is exclusive to first-time AWS users and expires after the initial 12-month window.

• Always-free tier (no time limit, usage-capped)
  Even during the 12-month new user period, and permanently after it, there's a usage-capped free tier for certain SageMaker resources. For your TensorFlow workflow, the key one is 50 hours per month of ml.t3.medium SageMaker Studio notebook instance usage. As long as you don't exceed this monthly limit, you won't be charged for this instance type—no expiration date attached.

Quick Tips to Avoid Unexpected Charges #

• Free tiers only apply to specific instance types. If you switch to higher-powered instances (like GPU-based ml.p3 or ml.g5 instances for faster TensorFlow training), you'll be charged at standard rates even if you're within the new user 12-month window.
• Keep tabs on your usage via the AWS Billing Dashboard—you can set up alerts to notify you when you're getting close to your free tier limits.
• For new users, the 12-month free tier and always-free tier can be used at the same time for eligible resources, just make sure you track each limit separately.

内容的提问来源于stack exchange，提问作者Huzo