Sure! Here’s the translation into American English:
—
Amazon has introduced an innovative on-demand deployment feature for custom models on its Amazon Bedrock platform. This new tool allows users to tailor versions of foundational models according to their specific needs, utilizing processes such as fine-tuning and distillation. With this functionality, custom models can be activated only when required, enabling real-time request processing without the need for pre-provisioned computing resources.
Another relevant feature of this launch is the token-based pricing model, which charges users based on the number of tokens used during inference. This “pay-as-you-go” approach complements the current provisioned performance option, providing users with the necessary flexibility to choose the deployment method that best fits their workload requirements and cost objectives.
The process for deploying custom models on Amazon Bedrock involves several stages, starting with defining the use case and preparing data. Users can then customize the model using the fine-tuning or distillation tools available on the platform. Once the model is adapted, the evaluation and deployment phase follows, where the on-demand deployment option plays a crucial role.
Users have two options for implementing their custom models: through the Amazon Bedrock console, which offers an intuitive and user-friendly interface, or via APIs and SDKs. Those opting for the console will find a guided process from model selection to deployment creation, which also allows for real-time monitoring of its status.
It’s important for users to consider certain operational considerations when utilizing this new feature. Factors such as latency, regional availability, and quota limitations can affect the effectiveness of the solution. Therefore, it is advisable to gain a solid understanding of these aspects and to apply appropriate cost management strategies.
If users decide not to continue using the on-demand deployment after evaluation, it is essential to clean up resources to avoid additional charges. This process can be easily executed through the console or by using the provided APIs.
The introduction of this deployment option underscores Amazon’s commitment to making artificial intelligence infrastructure more accessible and flexible, providing businesses with an optimized way to utilize custom models according to their needs. This advancement represents a significant step forward in optimizing costs, simplifying operations, and scaling according to variable usage patterns.
via: MiMub in Spanish
