ttnn.group_norm

ttnn.group_norm(input_tensor: ttnn.Tensor, *, memory_config: ttnn.MemoryConfig | None = None, epsilon: float, input_mask: ttnn.Tensor | None = None, weight: ttnn.Tensor | None = None, bias: ttnn.Tensor | None = None, dtype: ttnn.DataType | None = None, core_grid: CoreGrid | None = None, inplace: bool | None = True, output_layout: ttnn.Layout | None = None) → ttnn.Tensor

Compute group_norm over input_tensor.

Parameters:

input_tensor (ttnn.Tensor) – the input tensor.

Keyword Arguments:

• memory_config (ttnn.MemoryConfig, optional) – Memory configuration for the operation. Defaults to None.

• num_groups (int) –

• epsilon (float) – 1e-12.

• input_mask (ttnn.Tensor, optional) – Defaults to None.

• weight (ttnn.Tensor, optional) – Defaults to None.

• bias (ttnn.Tensor, optional) – Defaults to None.

• dtype (ttnn.DataType, optional) – Defaults to None.

• core_grid (CoreGrid, optional) – Defaults to None.

• inplace (bool, optional) – Defaults to True.

• output_layout (ttnn.Layout, optional) – Defaults to None.

Returns:

ttnn.Tensor – the output tensor.