LostTech.TensorFlow : API Documentation

Type EvalSpec

Namespace tensorflow_estimator.python.estimator.training

Parent PythonObjectContainer

Interfaces IEvalSpec

Public static methods

EvalSpec NewDyn(object input_fn, ImplicitContainer<T> steps, object name, object hooks, object exporters, ImplicitContainer<T> start_delay_secs, ImplicitContainer<T> throttle_secs)

Initialize an `LinearOperatorCirculant`.

This `LinearOperator` is initialized to have shape `[B1,...,Bb, N, N]` by providing `spectrum`, a `[B1,...,Bb, N0, N1, N2]` `Tensor` with `N0*N1*N2 = N`.

If `input_output_dtype = DTYPE`:

* Arguments to methods such as `matmul` or `solve` must be `DTYPE`. * Values returned by all methods, such as `matmul` or `determinant` will be cast to `DTYPE`.

Note that if the spectrum is not Hermitian, then this operator corresponds to a complex matrix with non-zero imaginary part. In this case, setting `input_output_dtype` to a real type will forcibly cast the output to be real, resulting in incorrect results!

If on the other hand the spectrum is Hermitian, then this operator corresponds to a real-valued matrix, and setting `input_output_dtype` to a real type is fine.
object input_fn
ImplicitContainer<T> steps
object name
A name to prepend to all ops created by this class.
object hooks
object exporters
ImplicitContainer<T> start_delay_secs
ImplicitContainer<T> throttle_secs

Public properties

object exporters get; set;

object hooks get; set;

object input_fn get; set;

object name get; set;

object PythonObject get;

object start_delay_secs get; set;

object steps get; set;

object throttle_secs get; set;