Microsoft Technology Licensing, LLC

Saurabh Mohan Kulkarni

Marc Tremblay

Tiyasa Mitra

Sujeeth Subramanya Bharadwaj

Manuel Costa

Stavros Volos

David Thomas Chisnall

Cédric Alain Marie Fournet

Samuel Alexander Webster

Andrew Wagner

Daniel John Pelham Wilkinson

Graham Bernard Cunningham

Bharadwaj Pudipeddi

Jinwen Xi

Richard William Osborne

Nuno Claudino Pereira Lopes

Maral Mesmakhosroshahi

Kapil Vaswani

Matthias Baenninger

Yimin Deng

Melur K Raghuraman

Andy Wagner

Bryan William Tuttle

Ho-Yuen Chau

Carlos Jose Cela

Layali Rashid

A peripheral device, for use with a host, comprises one or more compute elements a security module and at least one encryption unit. The security module is configured to form a trusted execution environment on the peripheral device for processing sensitive data using sensitive code. The sensitive data and sensitive code are provided by a trusted computing entity which is in communication with the host computing device. The at least one encryption unit is configured to encrypt and decrypt data transferred between the trusted execution environment and the trusted computing entity via the host computing device. The security module is configured to compute and send an attestation to the trusted computing entity to attest that the sensitive code is in the trusted execution environment.

Peripheral device

Embodiments of the present disclosure include an error recovery method comprising detecting a computing error, restarting a first artificial intelligence processor of a plurality of artificial intelligence processors processing a data set, and loading a model in the artificial intelligence processor, wherein the model corresponds to a same model processed by the plurality of artificial intelligence processors during a previous processing iteration by the plurality of artificial intelligence processors on data from the data set.

Systems and methods for error recovery

Systems and methods for pipelined neural network processing with continuous and asynchronous updates are described. A method for processing a neural network comprising L layers, where L is an integer greater than two, includes partitioning the L layers among a set of computing resources configured to process forward passes and backward passes associated with each of the L layers. The method further includes initiating processing of the forward passes and the backward passes using the set of computing resources. The method further includes upon completion of a first set of forward passes and a first set of backward passes associated with a first layer of the L layers, initiating update of parameters associated with the first layer when gradients are available for updating the parameters associated with the first layer without waiting to calculate gradients associated with any of remaining L layers.

Pipelined neural network processing with continuous and asynchronous updates

A method of increasing computer hardware efficiency of a matrix computation. The method comprises receiving at a computer processing device, digital signals encoding one or more operations of the matrix computation, each operation including one or more operands. The method further comprises, responsive to determining, by a sparse data check device of the computer processing machine, that an operation of the matrix computation includes all dense operands, forwarding the operation to a dense computation device of the computer processing machine configured to perform the operation of the matrix computation based on the dense operands. The method further comprises, responsive to determining, by the sparse data check device, that an operation of the matrix computation includes one or more sparse operands, forwarding the operation to a sparse computation device configured to perform the operation of the matrix computation.

Accelerator for dense and sparse matrix computations

Embodiments of the present disclosure include a method for token-position handling comprising: processing a first sequence of tokens to produce a second sequence of tokens, wherein the second sequence of tokens has a smaller number of tokens than the first sequence of tokens; masking at least some tokens in the second sequence to produce masked tokens; moving the masked tokens to the beginning of the second sequence to produce a third sequence; encoding tokens in the third sequence into a set of numeric vectors in a first array; and processing the first array in a transformer neural network to determine correlations among the third sequence, the processing the first array producing a second array.

Token-position handling for sequence based neural networks

Techniques for training neural networks are provided. According to one set of embodiments, a first array is processed in a spreading component to produce a second array, where a first dimension of the first array corresponds to at least one sequence of approximately orthogonal numeric vectors representing tokens, and where the spreading component combines values along the first dimension. The second array is processed in a transformer neural network to determine correlations between the sequence, which produces a third array. One or more batches of the third array are processed in a de-spreading component to produce a fourth array.

Spread neural networks

Methods for gradient accumulation with free momentum are performed by systems and devices during neural network model training. An accumulator that includes a processor circuit and a memory element generates free momentum between passes of a neural network model training process. The processor circuit receives a difference weight (gradient) and generates a first input by applying a weighting parameter thereto. The processor circuit obtains a prior weight from the memory element and generates a second input by applying another weighting parameter thereto. The processor circuit generates a filtered input with momentum by filtering the first and second input. The memory element generates a stored next pass weight by accumulating the filtered input with the prior weight. A computing resource then processes the next pass of the neural network model training using the stored next pass weight. The methods, systems, and devices are applicable to pipelined model parallelism training processes.

System and method for gradient accumulation with free momentum

A bare metal resource includes a trusted portion and an untrusted portion. The trusted portion includes trusted hardware, an image repository, and a clearance manager. The clearance manager is executable during bootup of the bare metal resource to perform a clearance process on the untrusted portion, including deleting the BIOS in the untrusted portion and loading a trusted BIOS from the image repository on the untrusted hardware, to place the untrusted portion in a trusted state. The bare metal resource may be provisioned to a tenant of a cloud provider after being placed in the trusted state.

Clearance of bare metal resource to trusted state usable in cloud computing

Reduced precision computer number formats inherently limit the quantity of discrete numeric values that can be represented. Therefore, the solution values of an arithmetic function, for each numeric value that is individually and uniquely expressible utilizing such a reduced precision computer number format, can be precomputed since the quantity of unique solution values can be limited to a quantity that can be conveniently stored, such as in an array. Subsequently, rather than computing the solution value of such an arithmetic function, for a given input value, the precomputed array can be referenced and a solution value corresponding to the given input value can be read from the array. Reading numeric values from an array can be substantially faster than computing solution values of a computationally-expensive arithmetic function.

Computing device performance of low precision arithmetic functions with arrays of pre-calculated values

Growing community of inventors

Redmond, WA, United States of America

Saurabh Mohan Kulkarni