mindspore.mint.nn.Linear

class mindspore.mint.nn.Linear(in_features, out_features, bias=True, weight_init=None, bias_init=None, dtype=None)[source]

The linear connected layer.

Applies linear connected layer for the input. This layer implements the operation as:

\[\text{outputs} = X * kernel + bias\]

Warning

On the Ascend platform, if bias is False , the x cannot be greater than 6D in PYNATIVE or KBK mode.

where \(X\) is the input tensors, \(\text{kernel}\) is a weight matrix with the same data type as the \(X\) created by the layer, and \(\text{bias}\) is a bias vector with the same data type as the \(X\) created by the layer (only if the parameter bias is True).

Warning

In PyNative mode, if bias is False , the x cannot be greater than 6D.

Parameters

in_features (int) – The number of features in the input space.
out_features (int) – The number of features in the output space.
bias (bool, optional) – Specify whether the layer use a bias vector \(\text{bias}\). Default: True.
weight_init (Union[Tensor, str, Initializer, numbers.Number], optional) – The trainable weight_init parameter. The dtype is same as x. The values of str refer to the function mindspore.common.initializer.initializer(). Default: None, weight will be initialized using HeUniform.
bias_init (Union[Tensor, str, Initializer, numbers.Number], optional) – The trainable bias_init parameter. The dtype is same as x. The values of str refer to the function mindspore.common.initializer.initializer(). Default: None, bias will be initialized using Uniform.
dtype (mindspore.dtype, optional) – Data type of parameter. Default: None.

Inputs:

x (Tensor) - Tensor of shape \((*, in\_features)\).

Outputs:

Tensor of shape \((*, out\_features)\).

Supported Platforms:

Ascend

Examples

>>> import mindspore
>>> x = mindspore.tensor([[180, 234, 154], [244, 48, 247]], mindspore.float32)
>>> net = mindspore.mint.nn.Linear(3, 4)
>>> output = net(x)
>>> print(output.shape)
(2, 4)