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 |
---|---|
|
NDArray-or-Symbol. The input. |
|
{‘float16’, ‘float32’, ‘float64’, ‘int32’, ‘int64’, ‘int8’, ‘uint8’}, required. Output data type. |
Value¶
out
The result mx.ndarray
Link to Source Code: http://github.com/apache/incubator-mxnet/blob/1.6.0/src/operator/tensor/amp_cast.cc#L37