Skip to content

pycauset.Float16Vector

A memory-mapped vector storing 16-bit floating point numbers (half precision). Inherits from VectorBase.

Constructor

pycauset.Float16Vector(n: int)
pycauset.Float16Vector(array: numpy.ndarray)

When constructed from a NumPy array, the array must be rank-1 with dtype float16.

Notes

Half precision is primarily a storage and bandwidth optimization. Supported operations are gated by the support matrix (see documentation/internals/DType System.md).