AWS sagemaker documentation change

Service: sagemaker · 2025-08-25 · Documentation low

File: sagemaker/latest/dg/nova-distillation.md

Summary

Updated section headers and added documentation about CloudWatch logs, successful training outputs, and sample configuration files

Security assessment

Added AWS KMS encryption example for S3 buckets which demonstrates security best practices, but does not address any specific security vulnerability

Diff

diff --git a/sagemaker/latest/dg/nova-distillation.md b/sagemaker/latest/dg/nova-distillation.md
index 1d509179c..543631e9e 100644
--- a//sagemaker/latest/dg/nova-distillation.md
+++ b//sagemaker/latest/dg/nova-distillation.md
@@ -5 +5 @@
-ConceptsPrerequisitesData augmentation setupStart a training jobValidating augmented data quality
+ConceptsPrerequisitesSet up data augmentationStart a training jobCloudWatch logsSuccessful trainingValidate augmented data quality
@@ -17 +17 @@ This quick start guide helps you get started with Amazon Nova model distillation
-  * Data augmentation setup
+  * Setting up data augmentation
@@ -19 +19,5 @@ This quick start guide helps you get started with Amazon Nova model distillation
-  * Start a SageMaker training job
+  * Starting a SageMaker training job
+
+  * CloudWatch logs
+
+  * Successful training
@@ -78 +82 @@ Mode distillation is particularly beneficial when:
-## Data augmentation setup
+## Setting up data augmentation
@@ -274 +278 @@ Example command to create an Amazon S3 bucket with AWS KMS encryption. Replace `
-## Start a SageMaker training job
+## Starting a SageMaker training job
@@ -417,0 +422,30 @@ The following sample notebook demonstrates how to run a SageMaker training job f
+## CloudWatch logs
+
+Logs are available in Amazon CloudWatch under the `/aws/sagemaker/TrainingJobs` log group in your AWS account. You will see one log file per host used for your training job.
+
+## Successful training
+
+For a successful training job, you will see the log message "Training is complete" at the end of the log.
+
+The output bucket contains the following files:
+
+  * `distillation_data/manifest.json`: Contains the location of augmented data. You can use this dataset to start an Amazon Nova fine-tuning job. Only SFT training is supported with this dataset.
+    
+        {
+      "distillation_data": "s3://customer_escrow_bucket/job_id/distillation_data/"
+    }                  
+
+  * `distillation_data/sample_training_data.jsonl`: This JSONL file contains 50 samples of augmented data for preview to help you determine data quality.
+
+  * `distillation_data/training_config.json`: This file contains recommended hyperparameters for Amazon Nova fine-tuning jobs. The following is an example file:
+    
+        {
+        "epochCount": 5, 
+        "learningRate": 1e-05, 
+        "batchSize": 1, 
+        "learningRateWarmupSteps": 1
+    }                  
+
+
+
+