AWS bedrock documentation change
Summary
Restructured documentation about model customization methods, replacing a comparison table with detailed bullet-point explanations for supervised fine-tuning, reinforcement fine-tuning, and distillation. Added specific usage instructions and links to related documentation pages.
Security assessment
The changes reorganize technical content about model training methods without addressing vulnerabilities, security incidents, or security features. The added content focuses on workflow improvements (e.g., using Lambda for reward functions) and documentation structure. No security-related terms, threat mitigations, or access controls were introduced or modified.
Diff
diff --git a/bedrock/latest/userguide/custom-models.md b/bedrock/latest/userguide/custom-models.md index d48523841..2910b5169 100644 --- a//bedrock/latest/userguide/custom-models.md +++ b//bedrock/latest/userguide/custom-models.md @@ -5,2 +4,0 @@ -Model customization methodsGuidelines for model customization - @@ -9 +7 @@ Model customization methodsGuidelines for model customization -Model customization is the process of providing training data to a model in order to improve its performance for specific use-cases. You can customize Amazon Bedrock foundation models in order to improve their performance and create a better customer experience. +Model customization is the process of providing training data to a model in order to improve its performance for specific use-cases. You can customize Amazon Bedrock foundation models in order to improve their performance and create a better customer experience. Amazon Bedrock currently provides the following customization methods. @@ -11 +9 @@ Model customization is the process of providing training data to a model in orde -## Model customization methods + * **Supervised fine-tuning** @@ -13 +11 @@ Model customization is the process of providing training data to a model in orde -Amazon Bedrock currently provides the following customization methods. +Provide _labeled_ data in order to train a model to improve performance on specific tasks. By providing a training dataset of labeled examples, the model learns to associate what types of outputs should be generated for certain types of inputs. The model parameters are adjusted in the process and the model's performance is improved for the tasks represented by the training dataset. @@ -15,5 +13 @@ Amazon Bedrock currently provides the following customization methods. -Customization method | Description | How to use | For more information ----|---|---|--- -Supervised fine-tuning | Provide _labeled_ data in order to train a model to improve performance on specific tasks. By providing a training dataset of labeled examples, the model learns to associate what types of outputs should be generated for certain types of inputs. The model parameters are adjusted in the process and the model's performance is improved for the tasks represented by the training dataset. | Create a training dataset of labeled input-output pairs that represent the specific tasks you want the model to improve on. Submit a fine-tuning job with your labeled dataset. | [Customize a model with fine-tuning in Amazon Bedrock](./custom-model-fine-tuning.html) -Reinforcement fine-tuning | Improves foundation model alignment with your specific use case through feedback-based learning. Instead of providing labeled input-output pairs, you define reward functions that evaluate response quality. The model learns iteratively by receiving feedback scores from these reward functions. | Use existing Bedrock invocation logs as training data or upload custom prompt datasets. Define reward functions using AWS Lambda to evaluate response quality. Amazon Bedrock automates the training workflow and provides real-time metrics to monitor model learning progress. | [Customize a model with reinforcement fine-tuning in Amazon Bedrock](./reinforcement-fine-tuning.html) -Distillation | Transfer knowledge from a larger more intelligent model (known as teacher) to a smaller, faster, and cost-efficient model (known as student). Amazon Bedrock automates the distillation process by using the latest data synthesis techniques to generate diverse, high-quality responses from the teacher model, and fine-tunes the student model. | Select a teacher model whose accuracy you want to achieve for your use case, and a student model to fine-tune. Then, provide use case-specific prompts as input data. Amazon Bedrock generates responses from the teacher model for the given prompts, and then uses the responses to fine-tune the student model. You can optionally provide labeled input data as prompt-response pairs. | [Customize a model with distillation in Amazon Bedrock](./model-distillation.html) +For more information about using supervised fine-tuning, see [Customize a model with fine-tuning in Amazon Bedrock](./custom-model-fine-tuning.html). @@ -21 +15 @@ Distillation | Transfer knowledge from a larger more intelligent model (known as -For information about model customization quotas, see [Amazon Bedrock endpoints and quotas](https://docs.aws.amazon.com/general/latest/gr/bedrock.html) in the AWS General Reference. After you customize a model, you can set up inference for the new custom model. For more information, see [Set up inference for a custom model](./model-customization-use.html). + * **Reinforcement fine-tuning** @@ -23 +17 @@ For information about model customization quotas, see [Amazon Bedrock endpoints -###### Note +Reinforcement fine-tuning improves foundation model alignment with your specific use case through feedback-based learning. Instead of providing labeled input-output pairs, you define reward functions that evaluate response quality. The model learns iteratively by receiving feedback scores from these reward functions. @@ -25 +19,9 @@ For information about model customization quotas, see [Amazon Bedrock endpoints -You are charged for model training based on the number of tokens processed by the model (number of tokens in training data corpus × number of epochs) and model storage charged per month per model. For more information, see [Amazon Bedrock pricing](https://aws.amazon.com/bedrock/pricing/). +You can upload your training prompt datasets or provide existing Bedrock invocation logs. You can define reward functions using AWS Lambda to evaluate response quality. Amazon Bedrock automates the training workflow and provides real-time metrics to monitor model learning progress. + +For more information about using reinforcement fine-tuning, see [Customize a model with reinforcement fine-tuning in Amazon Bedrock](./reinforcement-fine-tuning.html). + + * **Distillation** + +Use distillation to transfer knowledge from a larger more intelligent model (known as teacher) to a smaller, faster, and cost-efficient model (known as student). Amazon Bedrock automates the distillation process by using the latest data synthesis techniques to generate diverse, high-quality responses from the teacher model, and fine-tunes the student model. + +To use distillation, you select a teacher model whose accuracy you want to achieve for your use case, and a student model to fine-tune. Then, you provide use case-specific prompts as input data. Amazon Bedrock generates responses from the teacher model for the given prompts, and then uses the responses to fine-tune the student model. You can optionally provide labeled input data as prompt-response pairs. @@ -27 +29 @@ You are charged for model training based on the number of tokens processed by th -## Guidelines for model customization +For more information about using distillation see [Customize a model with distillation in Amazon Bedrock](./model-distillation.html). @@ -29 +30,0 @@ You are charged for model training based on the number of tokens processed by th -The ideal parameters for customizing a model depend on the dataset and the task for which the model is intended. You should experiment with values to determine which parameters work best for your specific case. To help, evaluate your model by running a model evaluation job. For more information, see [Evaluate the performance of Amazon Bedrock resources](./evaluation.html). @@ -31 +32,7 @@ The ideal parameters for customizing a model depend on the dataset and the task -Use the training and validation metrics from the [output files](./model-customization-analyze.html) generated when you [submit](./model-customization-submit.html) a model customization job to help you adjust your parameters. Find these files in the Amazon S3 bucket to which you wrote the output, or use the [GetCustomModel](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_GetCustomModel.html) operation. + + +For information about model customization quotas, see [Amazon Bedrock endpoints and quotas](https://docs.aws.amazon.com/general/latest/gr/bedrock.html) in the AWS General Reference. + +###### Note + +You are charged for model training based on the number of tokens processed by the model (number of tokens in training data corpus × number of epochs) and model storage charged per month per model. For more information, see [Amazon Bedrock pricing](https://aws.amazon.com/bedrock/pricing/). @@ -41 +48 @@ Tool use -Model customization access and security +Supervised fine-tuning