`analogvnn.fn.reduce_precision`#

Module Contents#

Functions#

`reduce_precision`(...)	Takes x and reduces its precision to precision by rounding to the nearest multiple of precision.
`stochastic_reduce_precision`(...)	Takes x and reduces its precision by rounding to the nearest multiple of precision with stochastic scheme.

analogvnn.fn.reduce_precision.reduce_precision(x: analogvnn.utils.common_types.TENSOR_OPERABLE, precision: analogvnn.utils.common_types.TENSOR_OPERABLE, divide: analogvnn.utils.common_types.TENSOR_OPERABLE) → analogvnn.utils.common_types.TENSOR_OPERABLE[source]#

Takes x and reduces its precision to precision by rounding to the nearest multiple of precision.

Parameters:

x (TENSOR_OPERABLE) – Tensor
precision (TENSOR_OPERABLE) – the precision of the quantization.
divide (TENSOR_OPERABLE) – the number of bits to be reduced

Returns:

TENSOR_OPERABLE with the same shape as x, but with values rounded to the nearest multiple of precision.

Return type:

TENSOR_OPERABLE

analogvnn.fn.reduce_precision.stochastic_reduce_precision(x: analogvnn.utils.common_types.TENSOR_OPERABLE, precision: analogvnn.utils.common_types.TENSOR_OPERABLE) → analogvnn.utils.common_types.TENSOR_OPERABLE[source]#

Takes x and reduces its precision by rounding to the nearest multiple of precision with stochastic scheme.