mindspore.nn.ClipByNorm

class mindspore.nn.ClipByNorm(axis=None)[source]

Clips tensor values to a maximum $L_{2}$ -norm.

The output of this layer remains the same if the $L_{2}$ -norm of the input tensor is not greater than the argument clip_norm. Otherwise the tensor will be normalized as:

output (X) = \frac{clip_norm * X}{L_{2} (X)},

where $L_{2} (X)$ is the $L_{2}$ -norm of $X$ .

Parameters: axis (Union[None, int, tuple(int)]) – Compute the L2-norm along the Specific dimension. Default: None, all dimensions to calculate.

Inputs:

x (Tensor) - Tensor of shape N-D. The type must be float32 or float16.
clip_norm (Tensor) - A scalar Tensor of shape $()$ or $(1)$ . Or a tensor shape can be broadcast to input x shape.

Outputs:

Tensor, clipped tensor with the same shape as the x, whose type is float32.

Raises

TypeError – If axis is not one of None, int, tuple.
TypeError – If dtype of x is neither float32 nor float16.

Supported Platforms:: Ascend GPU CPU

Examples

>>> net = nn.ClipByNorm()
>>> x = Tensor(np.random.randint(0, 10, [4, 16]), mindspore.float32)
>>> clip_norm = Tensor(np.array([100]).astype(np.float32))
>>> output = net(x, clip_norm)
>>> print(output.shape)
(4, 16)