Twitter, Inc.

Example computer-implemented methods, media, and systems for serving machine learning (ML) models using an asynchronous input/output (I/O) mechanism are disclosed. One example method includes receiving a first request for running a ML model to provide a first prediction. A first green thread is generated responsive to the first request and executed on an operating system (OS) thread to send a first asynchronous remote procedure call (RPC) to a multiple producer single consumer (MPSC) channel. A second request for running the ML model to provide a second prediction is received. A second green thread is generated responsive to the second request and executed on the OS thread to send a second RPC to the MPSC channel. The first and the second asynchronous RPCs are scheduled using a first and a second blocking threads respectively, which are used by the ML model to generate the first prediction and the second prediction.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for updating embeddings for user accounts on a social messaging platform. One of the methods includes receiving data having multiple data tuples, each data tuple indicating a user account and an item that the user account engaged with on the social messaging platform. A portion of the data is assigned to a respective multiple clusters. Embeddings are updated using an assigned cluster for each portion of the data. The updating includes: for each iteration step of multiple iteration steps, first embeddings are updated based on a first variable coefficient and a second variable coefficient. The updated first embeddings are stored into respective temporary data structures, which are then aggregated to generate aggregated first embeddings in a preserved data structure. Second embeddings are updated based on the aggregated first embeddings.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for managing cloud resources. One of the methods includes determining a computer resource to acquire from a cloud system for a first service, and first data comprising i) a first time period within which to acquire the computer resource and ii) a first urgency of the computer resource for the first service; posting, to a message queue, the first data; accessing, from the message queue, second data of a second agent comprising i) a second time period within which to acquire the computer resource for a second service, and ii) a second urgency of the computer resource for the second service; determining, whether to withdraw a first request for the computer resource; responsive to determining not withdraw the first request, determining an updated request; and upon acceptance of the updated request, acquiring the computer resource.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, in which a host client that has joined an conversation space within a social messaging platform receives user input; generates from the user input data representing a user input command to be executed by other user devices; and provides the data representing the user input command to the social messaging platform. The social messaging platform can provide second data representing the user input command to other user devices that have joined the conversation space. A client on a second user device can receive the data representing the user input command and a mixed audio stream, and render and output the mixed audio stream. That client can execute the user input command at the later of (i) the time location relative to the mixed audio stream, or (ii) when the second user device receives the second data.

According to an aspect, a method includes receiving, over a network, a conversation view request to retrieve messages from a conversation graph stored on a messaging platform from a first user account, generating a ranked list of messages for at least a portion of the messages from the conversation graph, identifying, from the ranked list of messages, a message authored by a second user account having an indirect relationship to the first user account in an engagement graph, selecting a social context object for the message based on a type of the indirect relationship, and transmitting, over the network, information to display the message and the social context object, the social context object providing information about the type of the indirect relationship.

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating a data processing pipeline. One of the methods includes receiving configuration data that identifies a) input data to be processed by a pipeline processing system that includes one or more subsystems and b) one or more data processing parameters; accessing two or more templates, wherein each template includes a set of data processing stages; selecting, from the two or more templates, one or more specific templates that have a plurality of data processing stages; generating the data processing pipeline that i) includes the plurality of data processing stages and ii) indicates one or more processing steps for the respective subsystem to perform on respective data; and causing the one or more subsystems to perform the processing steps to generate output data from the input data according to the one or more data processing parameters.

A sports training apparatus includes a flat mat having visible indicia adapted to indicate placement of at least one foot of a user of the apparatus, and a foot restraint adapted to hold a front foot of the user so that the front foot remains in contact with a top surface of the flat mat. The front foot is in an intended direction of a swing of a sports implement.

According to an aspect, a messaging system comprising at least one processor and a non-transitory computer-readable medium storing executable instructions that when executed by the at least one processor cause the at least one processor to obtain a system load metric associated with a messaging platform, compute a pruning factor based on the system load metric, rank messages of a conversation graph using a plurality of first signals to form an intermediate ranked list, prune the intermediate rank list according to the pruning factor to obtain a candidate subset of messages, rank the candidate subset of messages using a plurality of second signals to form a ranked list of messages, and transmit, over a network, information to render at least a portion of the ranked list on a client application.

This disclosure relates to a messaging system that can generate, store, and share groups of display columns associated with a user account, where each group may include a single display column or two or more display columns. A method may include rendering a first group of display columns of a messaging platform on a user interface associated with a first computing device, transmitting group creation data for creating a second group of display columns of the messaging platform, rendering a list identifying the first group of display columns and the second group of display columns, and rendering, in response to selection of the second group from the list, the second group of display columns on the user interface.

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium that receive, from a first client that has joined an audio conversation space of a social messaging platform, user interface presentation data that represents one or more audio tones of background audio for the audio conversation space. Background audio data representing the one or more audio tones of background audio for the audio conversation space can be generated from the user interface presentation. Conversation audio data can be received from one or more clients. A mixed audio stream can be generated and can include the conversation audio data received from the one or more clients and one or more other audio signals generated from the background audio data representing the background audio for the audio conversation space. The mixed audio stream can be presented to other client devices that have joined the audio conversation space.