glorot_normal¶

dragon.random.glorot_normal(
  shape,
  scale=2.0,
  mode='fan_in',
  dtype='float32',
  **kwargs
)[source]¶

Return a tensor initialized from the glorot normal distribution.

\[\text{out} \sim \mathcal{N}(0, \frac{scale}{\text{fan}}) \]

Parameters:

• shape (Sequence[Union[int, dragon.Tensor]]) – The tensor shape.
• mode ({'fan_in', 'fan_out', 'fan_avg'}, optional) – The mode to compute fans.
• scale (float, optional, default=2.0) – The scale factor to distribution.
• dtype (str, optional, default='float32') – The optional data type.

Returns:

dragon.Tensor – The output tensor.