Abstract: A method for artificial intelligence (AI) inferencing workload allocation includes, at a computing device of a distributed AI inferencing platform, receiving an estimated prompt load and an estimated generation load of an AI inferencing workload to be fulfilled by a processing unit of a computing node of the distributed AI inferencing platform. Based at least in part on the estimated prompt load and the estimated generation load, an inference unit (IU) processing load is estimated, the IU processing load to be applied to the processing unit while fulfilling the AI inferencing workload. Fractional processing capacity of the processing unit is allocated for fulfilling the AI inferencing workload based at least in part on the IU processing load.

Abstract: An online system receives multiple candidate components for including in content items to be presented to online system users. Upon identifying an opportunity to present content to a subject user of the online system, the online system dynamically generates an optimal content item for presentation to the subject user that includes one or more candidate components. Candidate components included in the optimal content item are associated with a predicted marginal effect on a performance metric associated the optimal content item. This marginal effect may be predicted using a machine-learned model that is trained using historical performance information about content items that were presented to viewing users of the online system having at least a threshold measure of similarity to the subject user and one or more features associated with candidate components included in these content items and in the optimal content item.

Abstract: An online system receives multiple candidate components for including in content items to be presented to online system users. Upon identifying an opportunity to present content to a subject user of the online system, the online system dynamically generates an optimal content item for presentation to the subject user that includes one or more candidate components. Candidate components included in the optimal content item are associated with a predicted marginal effect on a performance metric associated the optimal content item. This marginal effect may be predicted using a machine-learned model that is trained using historical performance information about content items that were presented to viewing users of the online system having at least a threshold measure of similarity to the subject user and one or more features associated with candidate components included in these content items and in the optimal content item.

Abstract: An online system provides feedback to a content provider creating a content item for a target audience. The feedback may include a score, recommendation, or error notification for a creative such as an image, video, or text to be included in the content item. The score indicates a likelihood that users of the online system will interact with the content item having the creative. Modifying the content item based on recommendations may result in a different score for the content item. The online system trains a machine learning model to generate the scores. The model learns which creatives are popular among particular audiences. The online system provides error notifications if the content item violates a rule. The online system can generate the content item even if there are rule violations. The feedback is displayed inline on a graphical user interface while the content provider is creating the content item.