Abstract: Some embodiments involve a reinforcement learning based framework for training a natural media agent to learn a rendering policy without human supervision or labeled datasets. The reinforcement learning based framework feeds the natural media agent a training dataset to implicitly learn the rendering policy by exploring a canvas and minimizing a loss function. Once trained, the natural media agent can be applied to any reference image to generate a series (or sequence) of continuous-valued primitive graphic actions, e.g., sequence of painting strokes, that when rendered by a synthetic rendering environment on a canvas, reproduce an identical or transformed version of the reference image subject to limitations of an action space and the learned rendering policy.

Abstract: Some embodiments involve a reinforcement learning based framework for training a natural media agent to learn a rendering policy without human supervision or labeled datasets. The reinforcement learning based framework feeds the natural media agent a training dataset to implicitly learn the rendering policy by exploring a canvas and minimizing a loss function. Once trained, the natural media agent can be applied to any reference image to generate a series (or sequence) of continuous-valued primitive graphic actions, e.g., sequence of painting strokes, that when rendered by a synthetic rendering environment on a canvas, reproduce an identical or transformed version of the reference image subject to limitations of an action space and the learned rendering policy.

Abstract: A non-photorealistic image rendering system and related techniques are described herein that train and implement machine learning models to reproduce digital images in accordance with various painting styles and constraints. The image rendering system can include a machine learning system that utilizes actor-critic based reinforcement learning techniques to train painting agents (e.g., models that include one or more neural networks) how to transform images into various artistic styles with minimal loss between the original images and the transformed images. The image rendering system can generate constrained painting agents, which correspond to painting agents that are further trained to reproduce images in accordance with one or more constraints. The constraints may include limitations of the color, width, size, and/or position of brushstrokes within reproduced images. These constrained painting agents may provide users with robust, flexible, and customizable non-photorealistic painting systems.

Abstract: The technology described herein is directed to a reinforcement learning based framework for training a natural media agent to learn a rendering policy without human supervision or labeled datasets. The reinforcement learning based framework feeds the natural media agent a training dataset to implicitly learn the rendering policy by exploring a canvas and minimizing a loss function. Once trained, the natural media agent can be applied to any reference image to generate a series (or sequence) of continuous-valued primitive graphic actions, e.g., sequence of painting strokes, that when rendered by a synthetic rendering environment on a canvas, reproduce an identical or transformed version of the reference image subject to limitations of an action space and the learned rendering policy.

Abstract: Various techniques are provided to support efficient data transfers over serial data streams. In one example, a serial device may be used to efficiently transfer data between a host device and the serial device over a data stream of a serial interface. A data stream value identifying the data stream may be stored in a register indexed by a tag associated with a command received from the host device. The command may be passed to a storage media device, wherein the passing is controlled by a processor of the serial device. The tag may be extracted from an address value received from the storage media device in response to execution of the command by the storage media device. The data stream value may be retrieved from the register using the extracted tag as an index without requiring an interrupt to the processor to determine the data stream value.

Abstract: Various methods, devices, systems, and machine readable mediums are provided which may be used in accordance with a protocol, such as a protocol used to support the transfer of data and commands between a host and a device. In one embodiment, a driver on a host device may be used to implement a protocol to provide support for various features of the universal serial bus (USB) attached storage (UAS) protocol when interfacing the host device with one or more connected devices that are compatible with either the USB 2.0 standard or the USB 3.0 standard. In this regard, a single driver on the host device may be used to support USB 2.0 and USB 3.0 devices. In another embodiment, such a protocol may be used to support USB 3.0 devices without requiring USB 3.0 streaming capabilities to be supported by host devices or connected devices.

Abstract: Systems and methods for a mass storage device attached to a host device use speculation about the host command likely to be received next from the host device based on a previously received command to improve throughput of accesses to the mass storage device. Host commands are used to speculatively produce commands for data storage devices of the mass storage device, such that host commands speculated as being likely next can be started during idle time of the data storage devices, based upon the probability that the speculation will be correct some of the time, and otherwise wasted idle time will be more efficiently used. Time taken by the host device to produce successive commands to the mass storage system is monitored, and future speculatively produced commands are parameterized to complete within the observed host time to produce new commands, making more efficient use of the data storage devices.

Abstract: Devices, methods, and systems for a mass storage device attached to a host device use speculation about the host command likely to be received next from the host device based on a previously received command to improve throughput of accesses to the mass storage device. Host device commands are used to speculatively produce commands for the data storage devices of the mass storage device, such that host commands speculated as being likely next host commands can be started during idle time of the data storage devices, based upon the probability that the speculation will be correct some of the time, and otherwise wasted idle time of the data storage devices will be more efficiently used. Time taken by the host device to produce successive commands to the mass storage system is monitored, and future speculatively produced commands are parameterized so that they complete within the observed host time to produce new commands, making more efficient use of the data storage devices of the mass storage system.

Abstract: Various methods, devices, systems, and machine readable mediums are provided which may be used in accordance with a protocol, such as a protocol used to support the transfer of data and commands between a host and a device. In one embodiment, a driver on a host device may be used to implement a protocol to provide support for various features of the universal serial bus (USB) attached storage (UAS) protocol when interfacing the host device with one or more connected devices that are compatible with either the USB 2.0 standard or the USB 3.0 standard. In this regard, a single driver on the host device may be used to support USB 2.0 and USB 3.0 devices. In another embodiment, such a protocol may be used to support USB 3.0 devices without requiring USB 3.0 streaming capabilities to be supported by host devices or connected devices.