mx.nd.amp.cast

Description

Cast function between low precision float/FP32 used by AMP.

It casts only between low precision float/FP32 and does not do anything for other types.

Arguments

Argument

Description

data

NDArray-or-Symbol.

The input.

dtype

{‘float16’, ‘float32’, ‘float64’, ‘int32’, ‘int64’, ‘int8’, ‘uint8’}, required.

Output data type.