Abstract: A server computer is connected to a plurality of client computers through a network, and controls objects in a Metaverse accessed by the client computers. The server computer includes a storage unit for storing an object ID specifying an object accessible in the Metaverse by the plurality of client computers and authenticity information associated with the object ID. The authenticity information indicates that the object is genuine. The server computer also includes a communication unit for communicating with each of the client computers. The server computer also includes an enquiry unit for causing the communication unit to transmit the authenticity information corresponding to the object ID to at least one of the plurality of client computers upon receipt of an enquiry request to enquire about the object ID of the object from one of the plurality of client computers.

Abstract: A computer-implemented method, a computer program product, and a computer processing system are provided for selecting from among multiple Graphics Processing Unit (GPU) execution modes for a Neural Network (NN) having a size greater than a threshold size. The multiple GPU execution modes include a normal memory mode, an Out-of-Core (OoC) execution mode, and a Unified Memory (UM) mode. The method includes starting an execution on the NN with the UM mode and measuring the memory usage for each of layers of the NN. The method further includes selecting an execution mode based on the memory usage of all of the layers.

Abstract: A method for converting a shape and a format of tensor data to meet a specific data format of a hardware accelerator is provided. The method receives input tensors L1 and L2, each being constants having a data format of < X x Y x Z >, and each further having an n-dimension input tensor shape as <Xn x Xn-1 x Xn-2 x ... x X1 >. The method stores input tensor shape. The method calculates an n-dimension modified shape of the input tensors by (a) setting a largest divisor of (Xn x Xn-1 x...x X1 ) ? L1 to S1, (b) setting a largest divisor of ((Xn x Xn-1 x...x X1 ) / S1) ? L2 to S2, (c) setting (((Xn x Xn-1 x... x X1 ) / (S1 x S2)) to S3, and (d) returning the n-dimension modified shape as < S3 x S2 x S1 >.

Abstract: A generated algorithm used by a neural network is captured during execution of an iteration of the neural network. A candidate algorithm is identified based on the generated algorithm. A determination is made that the candidate algorithm utilizes less memory than the generated algorithm. Based on the determination the neural network is updated by replacing the generated algorithm with the candidate algorithm.

Abstract: Verified snapshots are generated by obtaining, from one of a plurality of first nodes, a difference between a common data at a first time point and the common data at a second time point that is different from the first time point, generating a first snapshot of the common data at the first time point based on the difference, obtaining a hash of the common data at the first time point from one of the plurality of first nodes, and verifying the first snapshot at the first time point with the hash of the common data at the first time point.

Abstract: An example operation may include one or more of recording, by a snapshot node, a plurality of snapshots of a key-value storage based on a plurality of delta offsets from an initial snapshot of the key-value storage, receiving, by the snapshot node, an audit request from an audit node that contains an audit time, generating, by the snapshot node, a current snapshot based on an aggregation of the snapshots from the plurality of the snapshots up to a time of a transaction closest to the audit time, and executing, by the snapshot node, a chaincode based on a delta offset of the current snapshot from the time of the transaction to the audit time to restore a snapshot at the audit time.

Abstract: A method is presented for compressing data of a Rectified Linear Unit (ReLU) function on a graphical processing unit (GPU) employed in a learning process of a deep neural network. The method includes converting an initial data structure including nonzero data and zero data into a compressed data structure including only the nonzero data of the initial data structure as compressed data by generating a nonzero data bitmap region, generating a nonzero data number table region by employing a parallel reduction algorithm, calculating a nonzero data array index per block region of all blocks from the nonzero data number table region by employing a parallel prefix sum scan algorithm, allocating a buffer for the compressed data; and copying the nonzero data from the initial data structure into a nonzero data array region in a compressed data format in parallel.

Abstract: In a distributed processing system having multiple processing nodes including alive nodes and dead nodes, a method is provided for collecting an object from the alive nodes. The method includes maintaining a separate count value for each of remote nodes at which the object is remotely-referenced. The method further includes collecting the object for garbage collection when the separate count value for all of the remotes nodes is zero. The method also includes adding at least one per node sending counter responsive to a remote reference of the object being sent from a first remote node to a particular one of the remote nodes. The at least one per node sending counter is added at the first remote node to count a number of remote-references of the object being sent to the particular one of the remote nodes.

Abstract: A method is presented for compressing data of a Rectified Linear Unit (ReLU) function on a graphical processing unit (GPU) employed in a learning process of a deep neural network. The method includes converting an initial data structure including nonzero data and zero data into a compressed data structure including only the nonzero data of the initial data structure as compressed data by generating a nonzero data bitmap region, generating a nonzero data number table region by employing a parallel reduction algorithm, calculating a nonzero data array index per block region of all blocks from the nonzero data number table region by employing a parallel prefix sum scan algorithm, allocating a buffer for the compressed data; and copying the nonzero data from the initial data structure into a nonzero data array region in a compressed data format in parallel.

Abstract: Methods and systems for training a neural network include identifying units within a neural network, including a first unit for memory swapping and a second unit for re-computation to balance memory efficiency with computational efficiency. Each unit includes at least one layer of the neural network. Each unit has a first layer that is a checkpoint operation. During a feed-forward training stage, feature maps are stored in a first memory. The feature maps are output by the at least one layer of the first unit. The feature maps are swapped from the first memory to a second memory. During a backpropagation stage, the feature maps for the first unit are swapped from the second memory to the first memory. Feature maps for the second unit are re-computed.

Abstract: A method for reducing the cost of stack scanning in garbage collection (GC) includes, in the GC of the first-generation heap area, registering, in a nursery object reference list prepared for each thread, one or more addresses, within each stack, which each refer to a nursery object, and updating a scanning unnecessary area starting pointer such that the addresses listed in the nursery object reference list are included in the area from the bottom of the stack to the address pointed to by the scanning unnecessary area starting pointer. The method further includes, in the next GC of the first-generation heap area, for the area from the bottom of the stack to the address pointed to by the scanning unnecessary area starting pointer, performing the GC processing on the addresses included in the nursery object reference list.

Abstract: A server computer is connected to a plurality of client computers through a network, and controls objects in a Metaverse accessed by the client computers. The server computer includes a storage unit for storing an object ID specifying an object accessible in the Metaverse by the plurality of client computers and authenticity information associated with the object ID. The authenticity information indicates that the object is genuine. The server computer also includes a communication unit for communicating with each of the client computers. The server computer also includes an enquiry unit for causing the communication unit to transmit the authenticity information corresponding to the object ID to at least one of the plurality of client computers upon receipt of an enquiry request to enquire about the object ID of the object from one of the plurality of client computers.

Abstract: A server computer is connected to a plurality of client computers through a network, and controls objects in a Metaverse accessed by the client computers. The server computer includes a storage unit for storing an object ID specifying an object accessible in the Metaverse by the plurality of client computers and authenticity information associated with the object ID. The authenticity information indicates that the object is genuine. The server computer also includes a communication unit for communicating with each of the client computers. The server computer also includes an enquiry unit for causing the communication unit to transmit the authenticity information corresponding to the object ID to at least one of the plurality of client computers upon receipt of an enquiry request to enquire about the object ID of the object from one of the plurality of client computers.

Abstract: A method is provided for consistent data processing by first and second distributed processing systems having different data partitioning and routing mechanisms such that the first system is without states and the second system is with states. The method includes dividing data in each system into a same number of partitions based on a same key and a same hash function. The method includes mapping partitions between the systems in a one-to-one mapping. The mapping step includes calculating a partition ID based on the hash function and a total number of partitions, and dynamically mapping a partition in the first system to a partition in the second system, responsive to the partition in the first system being unmapped to the partition in the second system.

Abstract: Verified snapshots are generated by obtaining, from one of a plurality of first nodes, a difference between a common data at a first time point and the common data at a second time point that is different from the first time point, generating a first snapshot of the common data at the first time point based on the difference, obtaining a hash of the common data at the first time point from one of the plurality of first nodes, and verifying the first snapshot at the first time point with the hash of the common data at the first time point.

Abstract: Verified snapshots are generated by obtaining, from one of a plurality of first nodes, a difference between a common data at a first time point and the common data at a second time point that is different from the first time point, generating a first snapshot of the common data at the first time point based on the difference, obtaining a hash of the common data at the first time point from one of the plurality of first nodes, and verifying the first snapshot at the first time point with the hash of the common data at the first time point.

Abstract: Verified snapshots are generated by obtaining, from one of a plurality of first nodes, a difference between a common data at a first time point and the common data at a second time point that is different from the first time point, generating a first snapshot of the common data at the first time point based on the difference, obtaining a hash of the common data at the first time point from one of the plurality of first nodes, and verifying the first snapshot at the first time point with the hash of the common data at the first time point.

Abstract: A computer-implemented method is provided for managing GPU memory consumption by computational graph rewriting. The method includes constructing, by a hardware processor, a categorized topological ordering of a computational graph. The categorized topological ordering includes multiple computational nodes arranged in multiple levels. The method further includes estimating, by the hardware processor, the GPU memory consumption responsive to a level including two or more computational nodes from among the multiple computational nodes. The method also includes rewriting, by the hardware processor, the computational graph by linearizing the two or more computational nodes in the level to avoid overlapping of the GPU memory consumption by the two or more computational nodes responsive to the GPU memory consumption exceeding a threshold. The memory additionally includes managing the GPU memory consumption in accordance with the rewritten computational graph.

Abstract: An example operation may include one or more of recording, by a snapshot node, a plurality of snapshots of a key-value storage based on a plurality of delta offsets from an initial snapshot of the key-value storage, receiving, by the snapshot node, an audit request from an audit node that contains an audit time, generating, by the snapshot node, a current snapshot based on an aggregation of the snapshots from the plurality of the snapshots up to a time of a transaction closest to the audit time, and executing, by the snapshot node, a chaincode based on a delta offset of the current snapshot from the time of the transaction to the audit time to restore a snapshot at the audit time.

Abstract: A method is provided for consistent data processing by first and second distributed processing systems having different data partitioning and routing mechanisms such that the first system is without states and the second system is with states. The method includes dividing data in each system into a same number of partitions based on a same key and a same hash function. The method includes mapping partitions between the systems in a one-to-one mapping. The mapping step includes calculating a partition ID based on the hash function and a total number of partitions, and dynamically mapping a partition in the first system to a partition in the second system, responsive to the partition in the first system being unmapped to the partition in the second system.