The production inputs should be used to evaluate a new model.
Do not return its output to users yet.
Which method is most appropriate?
Mirror production inputs to the new model, record its output, and keep that output away from user responses and side effects.
Detailed explanation
Real distribution is evaluated safely.
Real distribution is evaluated safely.
Impact occurs before evaluation.
Impact occurs before evaluation.
Production distribution and boundaries differ.
Production distribution and boundaries differ.
The source of differences cannot be isolated.
The source of differences cannot be isolated.
Try it yourself
An example you can run in a temporary verification environment.
AWS公式AIF-C01 Domain 4.3の段階展開、評価、人的監督を確認する。Expected result
新モデルを利用者へ影響させず、同じ本番入力で比較評価できる。Key points
- Shadow deployment
- Same input
- Side-effect isolation
Notes
- Environment: AWS公式AIF-C01試験ガイドとAWS公式ドキュメントの確認
- Command output formatting can vary slightly by distribution or tool version.
- Run the example in a temporary directory or process when possible.
Foundation review
Read the scope first
Check whether the command acts on the current shell, a new process, an existing process, or a file.
Verify the observable result
Use the supplied command and compare the output with the expected result.