Deeploy.CommonExtensions.DataTypes.float16_t

class Deeploy.CommonExtensions.DataTypes.float16_t(value: int | float | Immediate, ctxt: _NetworkContext | None = None)

Bases: FloatImmediate

Methods

__init__(value: int | float | Immediate, ctxt: _NetworkContext | None = None)

`__init__`(value[, ctxt])
`checkPromotion`(value[, ctxt])	Checks whether a given Python-typed or Deeploy-typed value can be represented with the Deeploy type
`checkValue`(value[, ctxt])	This method tries to manually cast standard python's standard immediate float precision values (64 bits) to an arbitrary FP representation and check if the new representation is close enough to the original value.
`partialOrderUpcast`(otherCls)	This method checks whether a data type (cls) can be used to represent any value that can be represented by another data type (otherCls).

Attributes

`value`
`typeExponent`	Represents the number of bits reserved for the exponent part
`typeExponentMax`
`typeExponentOffset`
`typeMantissa`	Represents the number of bits reserved for the mantissa part
`typeMin`
`typeName`	The C typename of this type
`typeWidth`	the number of BITS to be assigned to the type

typeName: str = 'float16_t'

The C typename of this type

Type:: str

typeWidth: int = 16

the number of BITS to be assigned to the type

Type:: int

typeMantissa: int = 10

Represents the number of bits reserved for the mantissa part

Type:: int

typeExponent: int = 5

Represents the number of bits reserved for the exponent part

Type:: int

classmethod checkPromotion(value: _PythonType | Immediate, ctxt: _NetworkContext | None = None)

Checks whether a given Python-typed or Deeploy-typed value can be represented with the Deeploy type

Parameters:

value (Union[_PythonType, _DeeployType]) – Python-typed or Deeploy-typed value to be checked for promotion to cls
ctxt (Optional[_NetworkContext]) – Current NetworkContext

Returns:

Returns true if the value can be promoted to cls

Return type:

bool

classmethod checkValue(value: float | Iterable[float] | ndarray, ctxt: _NetworkContext | None = None): This method tries to manually cast standard python’s standard immediate float precision values (64 bits) to an arbitrary FP representation and check if the new representation is close enough to the original value.

classmethod partialOrderUpcast(otherCls: Type[Immediate]) → bool

This method checks whether a data type (cls) can be used to represent any value that can be represented by another data type (otherCls). For more information on partial order sets and type conversion, check:https://en.wikipedia.org/wiki/Partially_ordered_set https://en.wikipedia.org/wiki/Type_conversion

Parameters:: otherCls (Type[Immediate]) – The class you want to upcast an immediate of this cls to
Returns:: Returns true if this cls can be statically promoted to otherCls
Return type:: bool