AWS bedrock documentation change
Summary
Expanded documentation about model customization job outputs including training metrics, synthetic data artifacts, and prompt insights for distillation jobs. Added example S3 structure and improved parameter formatting.
Security assessment
Changes focus on improving documentation clarity and adding details about output artifacts. No security vulnerabilities or security features are mentioned. The updates relate to operational understanding rather than security controls.
Diff
diff --git a/bedrock/latest/userguide/model-customization-analyze.md b/bedrock/latest/userguide/model-customization-analyze.md index a0b58c704..721e5d7b4 100644 --- a//bedrock/latest/userguide/model-customization-analyze.md +++ b//bedrock/latest/userguide/model-customization-analyze.md @@ -7 +7,12 @@ -After a model customization job completes, you can analyze the results of the training process by looking at the files in the output S3 folder that you specified when you submitted the job or view details about the model. Amazon Bedrock stores your customized models in AWS-managed storage scoped to your account. +When your model customization job completes, you can analyze the results of the customization process. The following artifacts are uploaded to the S3 bucket that you specify when creating your model customization job: + + * **Training and validation metrics** – Amazon Bedrock provides training metrics for all model customization jobs. Validation metrics are also included with some model customization jobs. + + * **Synthetic data (Model Distillation only)** – Sample prompts from the synthetic dataset that Amazon Bedrock generated from your teacher model and used to fine tune your student model during the [distillation job](./submit-model-distillation-job.html). This information can help you further understand and validate how your custom model was trained. + + * **Prompt insights (Model Distillation only)** – A report of input prompts that were accepted and rejected (along with a reason why) during distillation. This information can help you fix and refine your prompts if you need to run another distillation job. + + + + +Amazon Bedrock stores your customized models in AWS managed storage scoped to your AWS account. @@ -11 +22 @@ You can also evaluate your model by running a model evaluation job. For more inf -The S3 output for a model customization job contains the following output files in your S3 folder. The validation artifacts only appear if you included a validation dataset. +The following example shows where you can fine training and validation metrics in an S3 bucket: @@ -25 +36 @@ The columns in the `step_wise_training_metrics.csv` file are as follows. - * step_number – The step in the training process. Starts from 0. + * `step_number` – The step in the training process. Starts from 0. @@ -27 +38 @@ The columns in the `step_wise_training_metrics.csv` file are as follows. - * epoch_number – The epoch in the training process. + * `epoch_number` – The epoch in the training process. @@ -29 +40 @@ The columns in the `step_wise_training_metrics.csv` file are as follows. - * training_loss – Indicates how well the model fits the training data. A lower value indicates a better fit. + * `training_loss` – Indicates how well the model fits the training data. A lower value indicates a better fit. @@ -31 +42 @@ The columns in the `step_wise_training_metrics.csv` file are as follows. - * perplexity – Indicates how well the model can predict a sequence of tokens. A lower value indicates better predictive ability. + * `perplexity` – Indicates how well the model can predict a sequence of tokens. A lower value indicates better predictive ability.