AWS nova documentation change
Summary
Restructured Nova Forge documentation: changed title, added new introduction, replaced detailed benefits section with bulleted capabilities list, removed explanations of reinforcement learning and responsible AI toolkit details, and updated section headers.
Security assessment
The changes are organizational and presentational without evidence of addressing specific vulnerabilities. The mention of 'custom safety guardrails' is maintained but reduced in detail. No security vulnerabilities, incidents, or weaknesses are referenced. The change focuses on feature descriptions rather than security improvements.
Diff
diff --git a/nova/latest/nova2-userguide/nova-forge.md b/nova/latest/nova2-userguide/nova-forge.md index 87b7b429f..aab74edd4 100644 --- a//nova/latest/nova2-userguide/nova-forge.md +++ b//nova/latest/nova2-userguide/nova-forge.md @@ -5 +5 @@ -Key benefits +# Amazon Nova Forge @@ -7 +7 @@ Key benefits -# Nova Forge +Amazon Nova Forge is a first-of-its-kind service that offers organizations the easiest and most cost-effective way to build their own frontier models using Nova. @@ -9 +9 @@ Key benefits -Nova Forge is the easiest and most cost-effective way to build your own frontier models using Amazon Nova. Customers can start their development from early model checkpoints, blend proprietary data with Amazon Nova-curated training data and host their custom models securely on AWS. +Amazon Nova Forge introduces the concept of “open training" models, which give organizations access to a variety of early model checkpoints and the ability to blend proprietary data with Amazon-curated data sets at every stage of model training. This allows the models to maximize learning from proprietary data while minimizing risk of forgetting foundational skills like reasoning. @@ -11 +11 @@ Nova Forge is the easiest and most cost-effective way to build your own frontier -## Key benefits +Nova Forge provides the following key capabilities: @@ -13 +13 @@ Nova Forge is the easiest and most cost-effective way to build your own frontier -### Setup access checkpoints across all phases of model development + * Access checkpoints across all phases of model development, and leverage new Nova models before they are widely available @@ -15 +15 @@ Nova Forge is the easiest and most cost-effective way to build your own frontier -Start your model development on SageMaker AI using early Nova checkpoints across pre-training, mid-training, or post-training phases. This lets you introduce your proprietary data at the optimal point in the model training, maximizing the model's learning from your data. + * Blend your proprietary data with Amazon Nova-curated training data @@ -17 +17 @@ Start your model development on SageMaker AI using early Nova checkpoints across -### Blend your proprietary data with Amazon Nova-curated training data + * Perform reinforcement learning with reward functions in your environment @@ -19 +19 @@ Start your model development on SageMaker AI using early Nova checkpoints across -Blend proprietary data with Amazon Nova-curated training data using Amazon provided SageMaker AI recipes. This approach lets you build a model that deeply understands your organization's proprietary knowledge, while minimizing risks like catastrophic forgetting and preserving foundational capabilities like reasoning. + * Use push-button recipes that are optimized to build with Nova through visual workflows or a command line interface @@ -21 +21 @@ Blend proprietary data with Amazon Nova-curated training data using Amazon provi -### Perform reinforcement learning with reward functions in your environment + * Use the built-in responsible AI toolkit to implement custom safety guardrails @@ -23 +22,0 @@ Blend proprietary data with Amazon Nova-curated training data using Amazon provi -Integrate reward functions in your environment for Reinforcement Fine Tuning (RFT). This allows the model to learn from feedback generated in your environment from your applications. @@ -25 +23,0 @@ Integrate reward functions in your environment for Reinforcement Fine Tuning (RF -### Use the built-in responsible AI toolkit to implement safety guardrails @@ -27,7 +24,0 @@ Integrate reward functions in your environment for Reinforcement Fine Tuning (RF -Use the responsible AI toolkit available in Nova Forge to configure the safety and content moderation settings of your custom model. You can adjust settings to meet your specific business needs in areas like safety, security and handling sensitive content. - -### Connect with experts at the AWS Generative AI Innovation Center - -Engage our generative AI experts through the Custom Model Program offered by the AWS Generative AI Innovation Center. Your team can work with experienced generative AI strategists, applied scientists and engineers to train and optimize models for your needs. - -For complete details on Nova Forge capabilities, setup and workflows, refer to the Nova Forge User Guide. @@ -41 +32 @@ To use the Amazon Web Services Documentation, Javascript must be enabled. Please -SDK reference +Limitations @@ -43 +34 @@ SDK reference -Responsible use +Nova Forge access and setup