Fujitsu Corporation

Atsushi Nukariya

Tsuyoshi Hashimoto

Takuya Sakamoto

Masahiro Doteguchi

Masaya Kato

A method including: determining a data type to be used for each of a plurality of operations, the plurality of operations being a series of operations to be used in machine learning processing; reducing a total processing time by modifying the determined data type to be used in each of the operations to a data type that is more accurate than the data type defined by the determining, the total processing time being a time period including an operation time for each of the plurality of operations and a conversion time taken to convert the data type; and executing each of the plurality of operations by using the modified data type of each of the plurality of operations.

Information processing device, machine learning method, and non-transitory computer-readable storage medium for storing machine learning program of converting data type so that efficiency of deep learning is improved

A non-transitory computer-readable recording medium stores therein a learning program for causing a computer to execute a process comprising: referring to, at time of learning a computation model that is a target of deep learning and has a plurality of nodes, a storage unit in which route information that indicates a calculation route followed by a tensor in each stage of learning prior to the time of learning, and statistical information regarding a position of a decimal point used in the calculation route are associated with each other; acquiring, when executing each piece of calculation processing set in each of the plurality of nodes at the time of learning, the statistical information corresponding to the route information that reaches each of the plurality of nodes; and executing the each piece of calculation processing using the position of the decimal point specified by the acquired statistical information.

Computer-readable recording medium recording learning program and learning method

An information processing apparatus includes: a memory; and a processor coupled to the memory and configured to: perform an arithmetic operation using an arithmetic operation target; repeat the arithmetic operation by using a calculated arithmetic operation result; obtain a ratio of, in a first number of elements which are included in the arithmetic operation result, a second number of elements in an expressible range as a predetermined-bit fixed point; and perform the arithmetic operation by using the predetermined-bit fixed point based on the ratio.

Selectively changing arithmetic data types used in arithmetic execution of deep learning applications based on expressible ratio and fluctuation value comparisons to threshold values

An information processing apparatus includes: a processor performs a scheduling process of scheduling a job for nodes and including: calculating, when one node executes a first job, a job execution end time when execution of the first job is completed by referring an execution history in which an execution time of a job is recorded; acquiring, from a load management node that manages a load of a metadata-process execution node which performing metadata processing to access metadata of a file among the nodes, the load of the metadata-process execution node at the job execution end time; and generating, when the load is equal to or more than a threshold, schedule data to cause a staging execution node which performs the metadata processing produced by staging, at the job execution end time, the metadata processing based on staging to a file having an execution result of the first job.

Information processing apparatus, stage-out processing method and recording medium recording job management program

A storage control apparatus configured to control a storage apparatus accessed by an information processing apparatus when the information processing apparatus executes a job, the storage control apparatus includes a memory configured to store plan information that indicates a plan in which the information processing apparatus executes the job and history information that indicates an history of access from the information processing apparatus to the storage apparatus when the information processing apparatus executes the job, and a processor coupled to the memory and configured to perform, based on the plan information and the history information, a prediction of whether the information processing apparatus accesses the storage apparatus in a certain time segment, and control power supply to the storage apparatus based on the prediction.

Storage control apparatus and method of controlling a storage apparatus based on prediction of storage access

An information processing system comprises a processor. The processor executes a process that causes the information processing system to perform first writing, when requested to write data, data from which to subtract remainder data obtained by dividing a data size of write target data by a first data size of first processing unit data, the first data size being a size of processing unit data of a reading/writing process in a first storage device, in the write target data requested to be written to the first storage device and second writing the remainder data not being written by the first writing to a second storage device in which to set a second data size of second processing unit data, the second data size being smaller than the first data size of the first processing unit data.

Information processing system and control method of information processing system

A storage control device includes a processor. The processor is configured to acquire a first size. The processor is configured to calculate, for each of a plurality of storage devices, a total size of first data stored in each of the plurality of storage devices. The first data satisfies a predetermined condition. The processor is configured to select the first data in an ascending order of the total size. A sum of the total size of the selected first data is the first size or less. The processor is configured to copy the selected first data from the plurality of storage devices to a first storage device different from the plurality of storage devices.

Growing community of inventors

Kawagoe, Japan

Atsushi Nukariya