Inference focus for offline training of SRAM inference engine in binary neural network
AssigneeInternational Business Machines Corporation
InventorsChia-Yu Chen, Jui-Hsin Lai, Ko-Tao Lee, Li-Wen Hung
Life of this patent
Estimating the term…
Abstract
A Static Random Access Memory (SRAM) device in a binary neural network is provided. The SRAM device includes an SRAM inference engine having an SRAM computation architecture with a forward path that include multiple SRAM cells. The multiple SRAM cells are configured to form a chain of SRAM cells such that an output of a given one of the multiple SRAM cells is an input to a following one of the multiple SRAM cells. The SRAM computation architecture is configured to compute a prediction from an input.
Claims19
A Static Random Access Memory (SRAM) device in a binary neural network, comprising:
- an SRAM inference engine having an SRAM computation architecture with a forward path that include multiple SRAM cells, wherein the multiple SRAM cells are configured to form a chain of SRAM cells such that an output of a given one of the multiple SRAM cells is an input to a following one of the multiple SRAM cells, the SRAM computation architecture being configured to compute a prediction from an input,
- wherein the multiple SRAM cells are arranged in an array to form a convolutional layer in the neural network.
A computer implemented method for inference training in a neural network, comprising:
- training an SRAM inference engine having an SRAM computation architecture with a forward path that include multiple SRAM cells, wherein the multiple SRAM cells are configured to form a chain of SRAM cells such that an output of a given one of the multiple SRAM cells is an input to a following one of the multiple SRAM cells, the SRAM computation architecture being configured to compute a prediction from an input,
- wherein the multiple SRAM cells are arranged in an array to form a convolutional layer in the neural network.
A computer program product for inference training in a neural network, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a computer to cause the computer to perform a method comprising:
- training an SRAM inference engine having an SRAM computation architecture with a forward path that include multiple SRAM cells, wherein the multiple SRAM cells are configured to form a chain of SRAM cells such that an output of a given one of the multiple SRAM cells is an input to a following one of the multiple SRAM cells, the SRAM computation architecture being configured to compute a prediction from an input,
- wherein the multiple SRAM cells are arranged in an array to form a convolutional layer in the neural network.
Dependencies are inferred from the claim text.