Systems, methods and devices for data quantization
AssigneeHuawei Technologies Co., Ltd.
InventorsManuel SALDANA, Barnaby DALTON, Vanessa COURVILLE
Life of this patent
Estimating the term…
Abstract
A memory control unit for handling data stored in a memory device includes a first interface to an interconnection with at least one memory bank; a second interface for communicating with a data requesting unit; and a memory quantization unit. The memory quantization unit is configured to: obtain, via the first interface, a first weight value from the at least one memory bank; quantize the first weight value to generate at least one quantized weight value having a shorter bit length than a bit length of the first weight value; and communicate the at least one quantized weight value to the data requesting unit via the second interface.
Claims22
A memory control unit comprising:
- a first interface to an interconnection with at least one memory bank;
- a second interface for communicating with a data requesting unit;
- and
- a memory quantization unit configured to:
- obtain, via the first interface, a first weight value from the at least one memory bank;
- quantize the first weight value to generate at least one quantized weight value having a shorter bit length than a bit length of the first weight value;
- and
- communicate the at least one quantized weight value to the data requesting unit via the second interface.
A method for quantizing data stored in a memory device, the method comprising:
- upon receiving, at a memory control unit, a quantized data request, obtaining a first weight value from the memory device;
- quantizing, with a memory quantization unit, the first weight value to generate at least one quantized weight value having a shorter bit length than a bit length of the first weight value;
- and
- communicating the at least one quantized weight value to a data requesting unit.
A circuit comprising:
- a memory control unit including a first interface to an interconnection with at least one memory bank;
- a second interface for communicating with a data requesting unit;
- and a memory quantization unit configured to:
- obtain, via the first interface, a first weight value from the at least one memory bank;
- quantize the first weight value to generate at least one quantized weight value having a shorter bit length than a bit length of the first weight value;
- and
- communicate the at least one quantized weight value to the data requesting unit via the second interface.
Dependencies are inferred from the claim text.