aliran tensor:: operasi:: ResourceApplyAdamWithAmsgrad

#include <training_ops.h>

Perbarui '*var' sesuai dengan algoritma Adam.

Ringkasan

$${lr}_t := {learning_rate} * {1 - ^t} / (1 - ^t)$$
$$m_t := * m_{t-1} + (1 - ) * g$$
$$v_t := * v_{t-1} + (1 - ) * g * g$$
$${v}_t := max{ {v}_{t-1}, v_t}$$
$${variable} := {variable} - {lr}_t * m_t / ({ {v}_t} + )$$

Argumen:

  • ruang lingkup: Objek Lingkup
  • var: Harus dari Variabel().
  • m: Harus dari Variabel().
  • v: Harus dari Variabel().
  • vhat: Harus dari Variabel().
  • beta1_power: Harus berupa skalar.
  • beta2_power: Harus berupa skalar.
  • lr: Faktor penskalaan. Pasti skalar.
  • beta1: Faktor momentum. Pasti skalar.
  • beta2: Faktor momentum. Pasti skalar.
  • epsilon: Istilah punggungan. Pasti skalar.
  • lulusan: Gradien.

Atribut opsional (lihat Attrs ):

  • use_locking: Jika True , pembaruan tensor var, m, dan v akan dilindungi oleh kunci; jika tidak, perilaku tersebut tidak terdefinisikan, namun mungkin menunjukkan lebih sedikit pertentangan.

Pengembalian:

Konstruktor dan Destruktor

ResourceApplyAdamWithAmsgrad (const :: tensorflow::Scope & scope, :: tensorflow::Input var, :: tensorflow::Input m, :: tensorflow::Input v, :: tensorflow::Input vhat, :: tensorflow::Input beta1_power, :: tensorflow::Input beta2_power, :: tensorflow::Input lr, :: tensorflow::Input beta1, :: tensorflow::Input beta2, :: tensorflow::Input epsilon, :: tensorflow::Input grad)
ResourceApplyAdamWithAmsgrad (const :: tensorflow::Scope & scope, :: tensorflow::Input var, :: tensorflow::Input m, :: tensorflow::Input v, :: tensorflow::Input vhat, :: tensorflow::Input beta1_power, :: tensorflow::Input beta2_power, :: tensorflow::Input lr, :: tensorflow::Input beta1, :: tensorflow::Input beta2, :: tensorflow::Input epsilon, :: tensorflow::Input grad, const ResourceApplyAdamWithAmsgrad::Attrs & attrs)

Atribut publik

operation

Fungsi publik

operator::tensorflow::Operation () const

Fungsi statis publik

UseLocking (bool x)

Struktur

tensorflow:: ops:: ResourceApplyAdamWithAmsgrad:: Attrs

Penyetel atribut opsional untuk ResourceApplyAdamWithAmsgrad .

Atribut publik

operasi

Operation operation

Fungsi publik

ResourceApplyAdamWithAmsgrad

 ResourceApplyAdamWithAmsgrad(
  const ::tensorflow::Scope & scope,
  ::tensorflow::Input var,
  ::tensorflow::Input m,
  ::tensorflow::Input v,
  ::tensorflow::Input vhat,
  ::tensorflow::Input beta1_power,
  ::tensorflow::Input beta2_power,
  ::tensorflow::Input lr,
  ::tensorflow::Input beta1,
  ::tensorflow::Input beta2,
  ::tensorflow::Input epsilon,
  ::tensorflow::Input grad
)

ResourceApplyAdamWithAmsgrad

 ResourceApplyAdamWithAmsgrad(
  const ::tensorflow::Scope & scope,
  ::tensorflow::Input var,
  ::tensorflow::Input m,
  ::tensorflow::Input v,
  ::tensorflow::Input vhat,
  ::tensorflow::Input beta1_power,
  ::tensorflow::Input beta2_power,
  ::tensorflow::Input lr,
  ::tensorflow::Input beta1,
  ::tensorflow::Input beta2,
  ::tensorflow::Input epsilon,
  ::tensorflow::Input grad,
  const ResourceApplyAdamWithAmsgrad::Attrs & attrs
)

operator::tensorflow::Operasi

 operator::tensorflow::Operation() const

Fungsi statis publik

Gunakan Penguncian

Attrs UseLocking(
  bool x
)