Custom model hyperparameters
The following reference content covers the hyperparameters that are available for training each Amazon Bedrock custom model.
A hyperparameter is a parameter that controls the training process, such as the learning rate or epoch count. You set hyperparameters for custom model training when you submit the fine tuning job with the Amazon Bedrock console or by calling the CreateModelCustomizationJob API operation. For guidelines on hyperparameter settings, see Guidelines for model customization.
Topics
- Amazon Titan text model customization hyperparameters
- Amazon Titan Image Generator G1 models customization hyperparameters
- Amazon Titan Multimodal Embeddings G1 customization hyperparameters
- Anthropic Claude 3 model customization hyperparameters
- Cohere Command model customization hyperparameters
- Meta Llama 2 model customization hyperparameters
- Meta Llama 3.1 model customization hyperparameters