Description
Currently, we return parameter values as floats. This choice is only enforced at the very end of sample as the computation is done using doubles but the output tensor are floats. For certain use cases, leaving the outputs as doubles may be desirable.
Acceptance criteria
Provide an option during the building to stage to specify whether the output tensor should return floats or doubles.
Description
Currently, we return parameter values as floats. This choice is only enforced at the very end of
sampleas the computation is done using doubles but the output tensor are floats. For certain use cases, leaving the outputs as doubles may be desirable.Acceptance criteria
Provide an option during the building to stage to specify whether the output tensor should return floats or doubles.