AWS sagemaker documentation change
Summary
Added link to recipe documentation, updated pre-training description from 'trillions of tokens' to 'tokens', and added clarification about post-training customization strategies (RFT/SFT).
Security assessment
Changes are editorial improvements and clarifications without security implications. The existing security statement about SageMaker-managed KMS encryption remains unchanged.
Diff
diff --git a/sagemaker/latest/dg/nova-model-training-job.md b/sagemaker/latest/dg/nova-model-training-job.md index c536bbf83..660123e20 100644 --- a//sagemaker/latest/dg/nova-model-training-job.md +++ b//sagemaker/latest/dg/nova-model-training-job.md @@ -11 +11 @@ Amazon SageMaker training jobs is an environment that enables you to train machi -The purpose of training is to customize the base Amazon Nova model using your proprietary data. The training process typically involves steps to prepare your data, choose a recipe, modify configuration parameters in YAML files, and submit a training job. The training process will output trained model checkpoint in a service-managed Amazon S3 bucket. You can use this checkpoint location for evaluation jobs. Nova customization on SageMaker training jobs stores model artifacts in a service-managed Amazon S3 bucket. Artifacts in the service-managed bucket are encrypted with SageMaker-managed KMS keys. Service-managed Amazon S3 buckets don't currently support data encryption using customer-managed KMS keys. +The purpose of training is to customize the base Amazon Nova model using your proprietary data. The training process typically involves steps to prepare your data, choose a [recipe](https://docs.aws.amazon.com/sagemaker/latest/dg/nova-model-recipes.html), modify configuration parameters in YAML files, and submit a training job. The training process will output trained model checkpoint in a service-managed Amazon S3 bucket. You can use this checkpoint location for evaluation jobs. Nova customization on SageMaker training jobs stores model artifacts in a service-managed Amazon S3 bucket. Artifacts in the service-managed bucket are encrypted with SageMaker-managed KMS keys. Service-managed Amazon S3 buckets don't currently support data encryption using customer-managed KMS keys. @@ -19 +19 @@ This section provides an overview of customization techniques and helps you choo -Large language model training consists of two major stages: pre-training and post-training. During pre-training, the model processes trillions of tokens of raw text and optimizes for next-token prediction. This process creates a pattern completer that absorbs syntax, semantics, facts, and reasoning patterns from web and curated text. However, the pre-trained model doesn't understand instructions, user goals, or context-appropriate behavior. It continues text in whatever style fits its training distribution. A pre-trained model autocompletes rather than follows directions, produces inconsistent formatting, and can mirror undesirable biases or unsafe content from the training data. Pre-training builds general competence, not task usefulness. +Large language model training consists of two major stages: pre-training and post-training. During pre-training, the model processes tokens of raw text and optimizes for next-token prediction. This process creates a pattern completer that absorbs syntax, semantics, facts, and reasoning patterns from web and curated text. However, the pre-trained model doesn't understand instructions, user goals, or context-appropriate behavior. It continues text in whatever style fits its training distribution. A pre-trained model autocompletes rather than follows directions, produces inconsistent formatting, and can mirror undesirable biases or unsafe content from the training data. Pre-training builds general competence, not task usefulness. @@ -24,0 +25,2 @@ Post-training transforms the pattern completer into a useful assistant. You run +In this section we will cover post training customization strategies: RFT and SFT. +