mindspore.ops.SparseApplyProximalAdagrad

class mindspore.ops.SparseApplyProximalAdagrad(use_locking=False)[源代码]

Updates relevant entries according to the proximal adagrad algorithm. Compared with ApplyProximalAdagrad, an additional index tensor is input.

\[\begin{split}\begin{array}{ll} \\ accum += grad * grad \\ \text{prox_v} = var - lr * grad * \frac{1}{\sqrt{accum}} \\ var = \frac{sign(\text{prox_v})}{1 + lr * l2} * \max(\left| \text{prox_v} \right| - lr * l1, 0) \end{array}\end{split}\]

Inputs of var, accum and grad comply with the implicit type conversion rules to make the data types consistent. If they have different data types, the lower priority data type will be converted to the relatively highest priority data type.

Parameters

use_locking (bool) – If true, the var and accum tensors will be protected from being updated. Default: False.

Inputs:

  • var (Parameter) - Variable tensor to be updated. The data type must be float16 or float32. The shape is \((N, *)\) where \(*\) means, any number of additional dimensions.

  • accum (Parameter) - Variable tensor to be updated, has the same shape and dtype as var.

  • lr (Union[Number, Tensor]) - The learning rate value, must be a float number or a scalar tensor with float16 or float32 data type.

  • l1 (Union[Number, Tensor]) - l1 regularization strength, must be a float number or a scalar tensor with float16 or float32 data type.

  • l2 (Union[Number, Tensor]) - l2 regularization strength, must be a float number or a scalar tensor with float16 or float32 data type.

  • grad (Tensor) - A tensor of the same type as var and grad.shape[1:] = var.shape[1:] if var.shape > 1.

  • indices (Tensor) - A tensor of indices in the first dimension of var and accum. If there are duplicates in indices, the behavior is undefined. Must be one of the following types: int32, int64 and indices.shape[0] = grad.shape[0].

Outputs:

Tuple of 2 tensors, the updated parameters.

  • var (Tensor) - The same shape and data type as var.

  • accum (Tensor) - The same shape and data type as accum.

Raises

  • TypeError – If use_locking is not a bool.

  • TypeError – If dtype of var, accum, lr, l1, l2 or grad is neither float16 nor float32.

  • TypeError – If dtype of indices is neither int32 nor int64.

  • RuntimeError – If the data type of var, accum and grad conversion of Parameter is not supported.

Supported Platforms:

Ascend GPU

Examples

>>> class Net(nn.Cell):
...     def __init__(self):
...         super(Net, self).__init__()
...         self.sparse_apply_proximal_adagrad = ops.SparseApplyProximalAdagrad()
...         self.var = Parameter(Tensor(np.array([[4.1, 7.2], [1.1, 3.0]], np.float32)), name="var")
...         self.accum = Parameter(Tensor(np.array([[0, 0], [0, 0]], np.float32)), name="accum")
...         self.lr = 1.0
...         self.l1 = 1.0
...         self.l2 = 0.0
...     def construct(self, grad, indices):
...         out = self.sparse_apply_proximal_adagrad(self.var, self.accum, self.lr, self.l1,
...                                                  self.l2, grad, indices)
...         return out
...
>>> net = Net()
>>> grad = Tensor(np.array([[1, 1], [1, 1]], np.float32))
>>> indices = Tensor(np.array([0, 1], np.int32))
>>> output = net(grad, indices)
>>> print(output)
(Tensor(shape=[2, 2], dtype=Float32, value=
[[ 2.09999990e+00,  5.19999981e+00],
 [ 0.00000000e+00,  1.00000000e+00]]), Tensor(shape=[2, 2], dtype=Float32, value=
[[ 1.00000000e+00,  1.00000000e+00],
 [ 1.00000000e+00,  1.00000000e+00]]))