Abstract: A computing system including a plurality of processing devices configured to execute a Mixture-of-Experts (MoE) layer included in an MoE model. The processing devices are configured to, in each of a plurality of iterations, at each of the processing devices, receive a respective plurality of input tokens. Executing the MoE layer further includes, at each of the processing devices, selecting one or more destination expert sub-models associated with the input tokens. Respective numbers k of expert sub-models selected differ across the iterations. At each of the processing devices, executing the MoE layer further includes conveying the input tokens to the one or more destination expert sub-models. Executing the MoE layer further includes generating one or more respective expert sub-model outputs at the one or more destination expert sub-models. Executing the MoE layer further includes generating and outputting an MoE layer output based on the one or more expert sub-model outputs.

Abstract: A method for preparing a rare earth anisotropic bonded magnetic powder, comprises the following steps: (1) preparing raw powder with RTBH as the main component, wherein, R is Nd or Pr/Nd, and T is a transition metal containing Fe; (2) adding La hydride or Ce hydride and copper powder to the raw powder to form a mixture; (3) subjecting the mixture to atmosphere diffusion heat treatment to give the rare earth anisotropic bonded magnetic powder.

Abstract: A computing system is provided, including a plurality of processing devices configured to execute a Mixture-of-Experts (MoE) layer included in an MoE model. The MoE layer includes a plurality of expert sub-models that each have a respective plurality of parameter values. The MoE layer is configured to be switchable between a data parallel mode and an expert-data-model parallel mode without conveying the respective parameter values of the expert sub-models among the plurality of processing devices.

Abstract: A computing system including a plurality of processing devices configured to execute a Mixture-of-Experts (MoE) layer included in an MoE model. The processing devices are configured to execute the MoE layer at least in part by, during a first collective communication phase between the processing devices, splitting each of a plurality of first input tensors along a first dimension to obtain first output tensors. Executing the MoE layer further includes processing the first output tensors at a respective a plurality of expert sub-models to obtain a plurality of second input tensors. Executing the MoE layer further includes, during a second collective communication phase between the processing devices, receiving the second input tensors from the expert sub-models and concatenating the second input tensors along the first dimension to obtain second output tensors. Executing the MoE layer further includes outputting the second output tensors as output of the MoE layer.

Abstract: The invention discloses a composite rare earth anisotropic bonded magnet and a preparation method thereof. The composite rare earth anisotropic bonded magnet comprises a Nd—Fe—B magnetic powder, a Sm—Fe—N magnetic powder, a binder and an inorganic nano-dispersant. The preparation method comprises steps of preparing a Nd—Fe—B magnetic powder by a HDDR method, preparing a Sm—Fe—N magnetic powder by a powder metallurgy method, mixing the Nd—Fe—B magnetic powder, the Sm—Fe—N magnetic powder, the binder and the inorganic nano-dispersant at a specific ratio to finally obtain the composite rare earth anisotropic bonded magnet. The invention, by adding an inorganic nano-dispersant, enables the full dispersion of the fine Sm—Fe—N powder during the mixing process of the binder, the Nd—Fe—B magnetic powder and the Sm—Fe—N powder, and thus makes the fine Sm—Fe—N powder and the binder evenly coated on the surface of the anisotropic Nd—Fe—B magnetic powder.

Abstract: The invention provides a lightweight hand-held stabilizer comprising a gimbal and a 7-shaped handle. The 7-shaped handle comprises a first connection arm (1) and a second connection arm (2), one end of the first connection arm (1) is rigidly connected to one end of the second connection arm (2), and a movable connector(s) is arranged at the end face of the other end of the first connection arm (1) and/or the second connection arm (2). The present invention can provide multiple using modes of the stabilizer handle, and the user experience of a stabilizer product is enhanced.

Abstract: Aspects of the present disclosure describe distributed fiber optic sensing (DFOS) systems, methods, and structures that employ a distributed fiber optic sensor placement procedure that advantageously provides a desirable sensor coverage over a network at minimal cost.

Abstract: Described is a novel framework, we call intent-based computing jobs assignment framework, for efficiently accommodating a clients' computing job requests in a mobile edge computing infrastructure. We define the intent-based computing job assignment problem, which jointly optimizes the virtual topology design and virtual topology mapping with the objective of minimizing the total bandwidth consumption. We use the Integer Linear Programming (ILP) technique to formulate this problem, and to facilitate the optimal solution. In addition, we employ a novel and efficient heuristic algorithm, called modified Steiner tree-based (MST-based) heuristic, which coordinately determines the virtual topology design and the virtual topology mapping. Comprehensive simulations to evaluate the performance of our solutions show that the MST-based heuristic can achieve an efficient performance that is close to the optimal performance obtained by the ILP solution.

Abstract: A method for preparing protein nanoparticle microspheres includes: dispersing a protein powder in water, adjusting a pH of a solution of the protein powder to between 10.0 and 12.0, stirring, centrifuging, and collecting a first upper dispersion; adjusting the first upper dispersion to be neutral using a cation exchange resin, centrifuging, to yield a protein nanoparticle dispersion; adding eugenol to the protein nanoparticle dispersion, stirring, centrifuging, and removing a precipitant, to yield a core-shell nanoprotein dispersion comprising a eugenol core and a protein shell; and adding the core-shell nanoprotein dispersion to ultrapure water, dialyzing, centrifuging, collecting a second upper dispersion, freeze-drying the second upper dispersion, to yield protein nanoparticle microspheres.

Abstract: A computing system including a plurality of processing devices configured to execute a Mixture-of-Experts (MoE) layer. The processing devices are configured to execute the MoE layer at least in part by receiving an input tensor including input tokens. Executing the MoE layer further includes computing a gating function output vector based on the input tensor and computing a sparse encoding of the input tensor and the gating function output vector. The sparse encoding indicates one or more destination expert sub-models. Executing the MoE layer further includes dispatching the input tensor for processing at the one or more destination expert sub-models, and further includes computing an expert output tensor. Executing the MoE layer further includes computing an MoE layer output at least in part by computing a sparse decoding of the expert output tensor. Executing the MoE layer further includes conveying the MoE layer output to an additional computing process.

Abstract: Embodiments are provided for facilitating multimodal extraction across multiple granularities. In one implementation, a set of features of a document for a plurality of granularities of the document is obtained. Via a machine learning model, the set of features of the document are modified to generate a set of modified features using a set of self-attention values to determine relationships within a first type of feature and a set of cross-attention values to determine relationships between the first type of feature and a second type of feature. Thereafter, the set of modified features are provided to a second machine learning model to perform a classification task.

Abstract: An anisotropic bonded magnet and a preparation method thereof are provided. By stacking magnets having different magnetic properties and/or densities, the magnets in the middle have high properties and the magnets at two ends and/or the periphery have low properties, thereby compensating for a property deviation caused by a difference in pressing densities during a pressing process, and improving the property uniformity of the magnets in an axial direction. The method solves the problem of “low in the middle and high at two ends” caused by the phenomenon of non-uniform magnetic field orientation and density along a height direction during orientation and densification.

Abstract: The present invention relates to an R-T-B sintered magnet and a preparation method thereof. The sintered magnet includes a grain boundary region T1, a shell layer region T2 and an R2Fe14B grain region T3; at 10 ?m to 60 ?m from a surface of the sintered magnet toward a center thereof, an area ratio of the shell layer region T2 to the R2Fe14B grain region T3 is 0.1 to 0.3, and a thickness of the shell layer region T2 is 0.5 ?m to 1.2 ?m; and an average coating percent of the shell layer region T2 on the R2Fe14B grain region T3 is 80% or more. In the present invention, by optimizing a preparation process and a microstructure of a traditional rare earth permanent magnet, diffusion efficiency of heavy rare earth in the magnet is improved, such that coercivity of the magnet is greatly improved, and manufacturing cost is reduced.

Abstract: The present disclosure discloses an yttrium (Y)-added rare-earth permanent magnetic material and a preparation method thereof. A chemical formula of the material expressed in atomic percentage is (YxRE1-x)aFebalMbNc, wherein 0.05?x?0.4, 7?a?13, 0?b?3, 5?c?20, and the balance is Fe, namely, bal=100-a-b-c; RE represents a rare-earth element Sm, or a combination of the rare-earth element Sm and any one or more elements of Zr, Nd and Pr; M represents Co and/or Nb; and N represents nitrogen. In the preparation method, the rare-earth element Y is utilized to replace the element Sm of a samarium-iron-nitrogen material. By regulating a ratio of the element Sm to the element Y, viscosity of an alloy liquid can be reduced, and an amorphous forming ability of the material is enhanced.

Abstract: This disclosure provides an accurate and personalized recommendation method based on a knowledge graph, which includes following steps: acquiring relevant knowledge of objects from a knowledge base according to historical behaviors of a user, and constructing a knowledge graph; initializing a vector representation of each node and its connection, and determining a receptive field of the node; generating training samples according to the historical behaviors of the user, and initializing a vector representation of all users and objects; acquiring a receptive field of an entity in the knowledge graph corresponding to the object in the training sample, then inputting the receptive field and the training sample to a graph neural network model to obtain predicted values of a possibility of an interaction between the user and the object.

Abstract: The present invention relates to a modified sintered Nd—Fe—B magnet, and a preparation method and a use thereof. The modified sintered Nd—Fe—B magnet is prepared by performing grain boundary diffusion on a matrix, wherein the matrix is a sintered Nd—Fe—B magnet; a grain boundary diffusion source for the grain boundary diffusion consists of a first diffusion source and a second diffusion source; the first diffusion source is a PrMx alloy, M being at least one selected from a group consisted of Cu, Al, Zn, Mg, Ga, Sn, Ag, Pb, Bi, Ni, Nb, Mn, Co, Fe, Ti, Cr, Zr, Mo and Ge; and the second diffusion source is heavy rare earth Dy and/or Tb.

Abstract: An anisotropic bonded magnet and a preparation method thereof are provided. Through a method of stacking magnets which are different in content of SmFeN and/or have different densities, the magnets in the middle have high properties and the magnets at two ends and/or the periphery have low properties, thereby compensating for a property deviation caused by a difference in densities during a pressing process, and improving the property uniformity of the magnets in an axial direction. The method avoids the phenomenon of non-uniform magnetic field orientation and density in a height direction during orientation and densification as well as the phenomenon of low in the middle and high at two ends.

Abstract: An anisotropic bonded magnet and a preparation method thereof are provided. By stacking magnets having different magnetic properties and/or densities, the magnets in the middle have high properties and the magnets at two ends and/or the periphery have low properties, thereby compensating for a property deviation caused by a difference in pressing densities during a pressing process, and improving the property uniformity of the magnets in an axial direction. The method solves the problem of “low in the middle and high at two ends” caused by the phenomenon of non-uniform magnetic field orientation and density along a height direction during orientation and densification.

Abstract: The present disclosure provides a visual navigation method and system for mobile devices based on QR code signposts. The visual navigation method for movement includes: autonomously planning an optimal route according to a destination specified by a user, performing positioning in an image captured by a trolley camera and recognizing signpost information, and adjusting a pose according to the signpost information and going to a signpost in the path. In this process, positioning is performed in real time and route planning is updated, and a user may change the destination at any time in a trolley navigation process according to a robust anomaly recovery decision. The method disclosed by the present disclosure not only proposes an indoor navigation solution that does not rely on a GPS signal and strong computation resources, but also can effectively meet an indoor scheduling scenario in which a path needs to be planned intelligently.