Python tensorflow.python.training.training_ops.apply_adam() Examples

The following are 15 code examples of tensorflow.python.training.training_ops.apply_adam(). You can vote up the ones you like or vote down the ones you don't like, and go to the original project or source file by following the links above each example. You may also want to check out all available functions/classes of the module tensorflow.python.training.training_ops , or try the search function .
Example #1
Source File: optimizer.py    From tensorflow-XNN with MIT License 6 votes vote down vote up 
def _apply_dense(self, grad, var):
        m = self.get_slot(var, "m")
        v = self.get_slot(var, "v")
        return training_ops.apply_adam(
            var,
            m,
            v,
            math_ops.cast(self._beta1_power, var.dtype.base_dtype),
            math_ops.cast(self._beta2_power, var.dtype.base_dtype),
            math_ops.cast(self._lr_t, var.dtype.base_dtype),
            math_ops.cast(self._beta1_t, var.dtype.base_dtype),
            math_ops.cast(self._beta2_t, var.dtype.base_dtype),
            math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
            grad,
            use_locking=self._use_locking,
            use_nesterov=True).op
Example #2
Source File: multistep_with_adamoptimizer.py    From tensor2tensor with Apache License 2.0 6 votes vote down vote up 
def _apply_cond(self, apply_fn, grad, var, *args, **kwargs):
    """Apply conditionally if counter is zero."""
    grad_acc = self.get_slot(var, "grad_acc")

    def apply_adam(grad_acc, apply_fn, grad, var, *args, **kwargs):
      total_grad = (grad_acc + grad) / tf.cast(self._n_t, grad.dtype)
      adam_op = apply_fn(total_grad, var, *args, **kwargs)
      with tf.control_dependencies([adam_op]):
        grad_acc_to_zero_op = grad_acc.assign(
            tf.zeros_like(grad_acc), use_locking=self._use_locking)
      return tf.group(adam_op, grad_acc_to_zero_op)

    def accumulate_gradient(grad_acc, grad):
      assign_op = tf.assign_add(grad_acc, grad, use_locking=self._use_locking)
      return tf.group(assign_op)  # Strip return value

    return tf.cond(
        tf.equal(self._get_iter_variable(), 0),
        lambda: apply_adam(grad_acc, apply_fn, grad, var, *args, **kwargs),
        lambda: accumulate_gradient(grad_acc, grad))
Example #3
Source File: multistep_with_adamoptimizer.py    From tensor2tensor with Apache License 2.0 6 votes vote down vote up 
def _apply_dense_in_action(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    beta1_power, beta2_power = self._get_beta_accumulators()
    return training_ops.apply_adam(
        var,
        m,
        v,
        tf.cast(beta1_power, var.dtype.base_dtype),
        tf.cast(beta2_power, var.dtype.base_dtype),
        tf.cast(self._lr_t, var.dtype.base_dtype),
        tf.cast(self._beta1_t, var.dtype.base_dtype),
        tf.cast(self._beta2_t, var.dtype.base_dtype),
        tf.cast(self._epsilon_t, var.dtype.base_dtype),
        grad,
        use_locking=self._use_locking).op
Example #4
Source File: optimizer.py    From BERT with Apache License 2.0 6 votes vote down vote up 
def _apply_dense(self, grad, var):
        m = self.get_slot(var, "m")
        v = self.get_slot(var, "v")
        return training_ops.apply_adam(
            var,
            m,
            v,
            math_ops.cast(self._beta1_power, var.dtype.base_dtype),
            math_ops.cast(self._beta2_power, var.dtype.base_dtype),
            math_ops.cast(self._lr_t, var.dtype.base_dtype),
            math_ops.cast(self._beta1_t, var.dtype.base_dtype),
            math_ops.cast(self._beta2_t, var.dtype.base_dtype),
            math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
            grad,
            use_locking=self._use_locking,
            use_nesterov=True).op
Example #5
Source File: nadam.py    From BERT with Apache License 2.0 6 votes vote down vote up 
def _apply_dense(self, grad, var):
        m = self.get_slot(var, "m")
        v = self.get_slot(var, "v")
        return training_ops.apply_adam(
            var,
            m,
            v,
            math_ops.cast(self._beta1_power, var.dtype.base_dtype),
            math_ops.cast(self._beta2_power, var.dtype.base_dtype),
            math_ops.cast(self._lr_t, var.dtype.base_dtype),
            math_ops.cast(self._beta1_t, var.dtype.base_dtype),
            math_ops.cast(self._beta2_t, var.dtype.base_dtype),
            math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
            grad,
            use_locking=self._use_locking,
            use_nesterov=True).op
Example #6
Source File: optimizer.py    From tensorflow-DSMM with MIT License 6 votes vote down vote up 
def _apply_dense(self, grad, var):
        m = self.get_slot(var, "m")
        v = self.get_slot(var, "v")
        return training_ops.apply_adam(
            var,
            m,
            v,
            math_ops.cast(self._beta1_power, var.dtype.base_dtype),
            math_ops.cast(self._beta2_power, var.dtype.base_dtype),
            math_ops.cast(self._lr_t, var.dtype.base_dtype),
            math_ops.cast(self._beta1_t, var.dtype.base_dtype),
            math_ops.cast(self._beta2_t, var.dtype.base_dtype),
            math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
            grad,
            use_locking=self._use_locking,
            use_nesterov=True).op
Example #7
Source File: nadam.py    From tensorflow-DSMM with MIT License 6 votes vote down vote up 
def _apply_dense(self, grad, var):
        m = self.get_slot(var, "m")
        v = self.get_slot(var, "v")
        return training_ops.apply_adam(
            var,
            m,
            v,
            math_ops.cast(self._beta1_power, var.dtype.base_dtype),
            math_ops.cast(self._beta2_power, var.dtype.base_dtype),
            math_ops.cast(self._lr_t, var.dtype.base_dtype),
            math_ops.cast(self._beta1_t, var.dtype.base_dtype),
            math_ops.cast(self._beta2_t, var.dtype.base_dtype),
            math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
            grad,
            use_locking=self._use_locking,
            use_nesterov=True).op
Example #8
Source File: adam.py    From lambda-packs with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking).op
Example #9
Source File: nadam_optimizer.py    From lambda-packs with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking,
        use_nesterov=True).op
Example #10
Source File: adam.py    From auto-alt-text-lambda-api with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking).op
Example #11
Source File: training_ops_test.py    From deep_image_model with Apache License 2.0 5 votes vote down vote up 
def _testTypesForAdam(self, var, m, v, grad, use_gpu):
    self.setUp()
    with self.test_session(use_gpu=use_gpu):
      var_t = variables.Variable(var)
      m_t = variables.Variable(m)
      v_t = variables.Variable(v)

      t = 1
      beta1 = np.array(0.9, dtype=var.dtype)
      beta2 = np.array(0.999, dtype=var.dtype)
      beta1_power = beta1**t
      beta2_power = beta2**t
      lr = np.array(0.001, dtype=var.dtype)
      epsilon = np.array(1e-8, dtype=var.dtype)
      beta1_t = constant_op.constant(beta1, self._toType(var.dtype), [])
      beta2_t = constant_op.constant(beta2, self._toType(var.dtype), [])
      beta1_power_t = variables.Variable(beta1_power)
      beta2_power_t = variables.Variable(beta2_power)
      lr_t = constant_op.constant(lr, self._toType(var.dtype), [])
      epsilon_t = constant_op.constant(epsilon, self._toType(var.dtype), [])
      variables.global_variables_initializer().run()

      self.assertAllCloseAccordingToType(var, var_t.eval())
      new_var, _, _ = self._adamUpdateNumpy(var, grad, t, m, v,
                                            lr, beta1, beta2, epsilon)
      apply_adam = training_ops.apply_adam(var_t, m_t, v_t, beta1_power_t,
                                           beta2_power_t, lr_t,
                                           beta1_t, beta2_t, epsilon_t, grad)
      out = apply_adam.eval()
      self.assertShapeEqual(out, apply_adam)
      self.assertAllCloseAccordingToType(new_var, out)
Example #12
Source File: adam.py    From deep_image_model with Apache License 2.0 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking).op
Example #13
Source File: optimizer.py    From NNCF with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v, self._beta1_power, self._beta2_power,
        self._lr_t, self._beta1_t, self._beta2_t,
        self._epsilon_t, grad, use_locking=self._use_locking).op
Example #14
Source File: adam.py    From Serverless-Deep-Learning-with-TensorFlow-and-AWS-Lambda with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking).op
Example #15
Source File: adam.py    From keras-lambda with MIT License 5 votes vote down vote up 
def _apply_dense(self, grad, var):
    m = self.get_slot(var, "m")
    v = self.get_slot(var, "v")
    return training_ops.apply_adam(
        var, m, v,
        math_ops.cast(self._beta1_power, var.dtype.base_dtype),
        math_ops.cast(self._beta2_power, var.dtype.base_dtype),
        math_ops.cast(self._lr_t, var.dtype.base_dtype),
        math_ops.cast(self._beta1_t, var.dtype.base_dtype),
        math_ops.cast(self._beta2_t, var.dtype.base_dtype),
        math_ops.cast(self._epsilon_t, var.dtype.base_dtype),
        grad, use_locking=self._use_locking).op