The present disclosure includes systems, methods, and computer-readable mediums for discovering capabilities of a hardware (HW) accelerator card. A processor may communicate a request for a listing of acceleration services to a HW accelerator card connected to the processor via the communication interface. The HW accelerator card may retrieve the listing from memory and provide a response to the processor that includes a listing of the HW acceleration services provided by the HW accelerator card.

A method, a device, and a computer program product for resource scheduling is disclosed. The method includes determining a job initiated by a virtual machine. The job requests to invoke at least one virtual function in a set of virtual functions associated with the virtual machine and each virtual function in the set of virtual functions is configured to utilize an accelerator resource to provide a single type of acceleration service. The method further includes determining, based on a job type of the job, a first accelerator resource allocated to the at least one virtual function. The accelerator resources required by the virtual functions invoked by the job may then be guaranteed, improving the execution efficiency of the job.

Remote access for debugging or profiling a remotely executing neural network graph can be performed by a client using an in-band application programming interface (API). The client can provide indicator flags for debugging or profiling in an inference request sent to a remote server computer executing the neural network graph using the API. The remote server computer can collect metadata for debugging or profiling during the inference operation using the neural network graph and send it back to the client using the same API. Additionally, the metadata can be collected at various granularity levels also specified in the inference request.

A method and apparatus for differentially optimizing a quality of service (QoS) includes: establishing a system model of a multi-task unloading framework; acquiring a mode for users executing a computation task, executing, according to the mode for users executing the computation task, the system model of the multi-task unloading framework; and optimizing a quality of service (QoS) on the basis of a multi-objective optimization method for a multi-agent deep reinforcement learning. According to the present invention, an unloading policy is calculated on the basis of a multi-user differentiated QoS of a multi-agent deep reinforcement learning, and with the differentiated QoS requirements among different users in a system being considered, a global unloading decision is performed according to a task performance requirement and a network resource state, and differentiated performance optimization is performed on different user requirements, thereby effectively improving a system resource utilization rate and a user service quality.

In a symmetric hardware accelerator system, an initial hardware accelerator is selected for an upgrade of firmware. The initial and other hardware accelerators handle workloads that have been balanced across the hardware accelerators. Workloads are rebalanced by directing workloads having low CPU utilization to the initial hardware accelerator. A CPU fallback is conducted of the workloads of the initial hardware accelerator to the CPU. While the CPU is handling the workloads, firmware of the initial hardware accelerator is upgraded.

Methods and apparatus to implement always-on context sensor hubs for processing multiple different types of data inputs are disclosed. An examples apparatus includes a first processor core to implement a host controller, and a second processor core to implement an offload engine. The host controller includes first logic to process sensor data associated with an electronic device when the electronic device is in a low power mode. The host controller is to offload a computational task associated with the sensor data to the offload engine. The offload engine includes second logic to execute the computational task.

A computing resource sharing system and a computing resource sharing method are provided. The method includes: in response to receiving a resource request signal from a resource request device, obtaining a foreground process, a background process, a name of a software service, and an operating status of the software service of a resource sharing device; determining a specific graphic computing resource to be shared according to the foreground process, the background process, the name of the software service, and the operating status of the software service; applying the specific graphic computing resource to assist the resource request device in performing a graphic computing operation; transmitting a graphic computing result of the graphic computing operation back to the resource request device.

The subject technology provides for dynamic task allocation for neural network models. The subject technology determines an operation performed at a node of a neural network model. The subject technology assigns an annotation to indicate whether the operation is better performed on a CPU or a GPU based at least in part on hardware capabilities of a target platform. The subject technology determines whether the neural network model includes a second layer. The subject technology, in response to determining that the neural network model includes a second layer, for each node of the second layer of the neural network model, determines a second operation performed at the node. Further the subject technology assigns a second annotation to indicate whether the second operation is better performed on the CPU or the GPU based at least in part on the hardware capabilities of the target platform.

The present disclosure provides a method and an apparatus for controlling a robot, a method and an apparatus for providing service and an electronic device. The method comprises: creating, by a local robot, a file system snapshot of an application in a local operating system, and synchronizing file system data of the application to a cloud robot; running, by the cloud robot, on a cloud virtual machine pre-running the same operating system as that of the local robot, the same application as that of the local robot (101); and reversely synchronizing a running result of the application to the local robot (102).

A method for securing Secure Objects that are protected from other software on a heterogeneous data processing system including a plurality of different types of processors wherein different portions of a Secure Object may run on different types of processors. A Secure Object may begin execution on a first processor then, depending on application requirements, the Secure Object may make a call to a second processor passing information to the second processor using a special inter-processor function call. The second processor performs the requested processing and then performs an inter-processor “function return” returning information as appropriate to the Secure Object on the first processor.