Computing system, server and on-vehicle device
A computing system includes a server, and an on-vehicle device mounted on a vehicle. The server includes: a server storage part containing a learned model; a simplification part which generates a contraction information for determination of calculation precision by using the learned model and an object of inference; and a server communication part for transmitting the contraction information to the on-vehicle device. The on-vehicle device includes: a signal input part to which an output signal from a sensor mounted on the vehicle is inputted; an on-vehicle communication part for receiving the contraction information; an inference part for making an inference on the output signal and a reconfigurable logic circuit; and a reconfiguration part for configuring the inference part in the logic circuit based on the contraction information.
1. A computing system comprising a server and an on-vehicle device mounted on a vehicle, wherein
the server includes:
a server storage containing a learned model;
a first processor which generates a contraction information for determination of calculation precision by using the learned model and an object of inference; and
a server communication circuitry for transmitting the contraction information to the on-vehicle device, and
the on-vehicle device includes:
a signal input to which an output signal from a sensor mounted on the vehicle is inputted;
an on-vehicle communication circuitry for receiving the contraction information;
an inference logic circuit for making an inference on the output signal;
a reconfigurable logic circuit; and
a reconfiguration circuitry for configuring the inference logic circuit based on the contraction information.
2. The computing system according to claim 1 , wherein
the on-vehicle device further includes an on-vehicle storage containing the learned model, and
the reconfiguration circuitry configures the inference logic circuit based on the learned model and the contraction information.
3. The computing system according to claim 1 , wherein
the first processor creates a low-precision model integrating the contraction information and the learned model,
the server communication circuitry transmits the low-precision model to the on-vehicle device, and
the reconfiguration circuitry configures the inference logic circuit by using the received low-precision model.
4. The computing system according to claim 1 , wherein
the learned model comprises linear components alone.
5. The computing system according to claim 1 , wherein
the learned model comprises multiple layers, and
the contraction information determines a precision of calculation for each of the multiple layers.
6. The computing system according to claim 1 , wherein
the first processor makes calculations with a reference precision and with a precision lower than the reference precision by using the learned model and the object of inference so as to define, as the contraction information, the lowest precision that provides a difference equal to or less than a predetermined value between a result of the calculation with the reference precision and a result of the calculation with the precision lower than the reference precision.
7. The computing system according to claim 1 , wherein
the contraction information is digit numbers of a mantissa part and an exponent part of a floating-point type.
8. An on-vehicle device mounted on a vehicle comprising:
a signal input to which an output signal from a sensor mounted on the vehicle is inputted;
an on-vehicle communication circuitry for receiving a contraction information for determination of calculation precision;
an inference logic circuit for making an inference on the output signal;
a reconfigurable logic circuit; and
a reconfiguration circuitry for configuring the inference logic circuit based on the contraction information.