returnn.frontend.const

Constant / full / fill / zeros / ones, etc

returnn.frontend.const.full(*, dims: Sequence[Dim], fill_value: int | float | complex | number | ndarray | bool | str | Tensor, dtype: str | None = None, device: str | None = None, sparse_dim: Dim | None = None, feature_dim: Dim | None = None) → Tensor

full, fill, constant.

https://data-apis.org/array-api/latest/API_specification/generated/array_api.full.html

Also see convert_to_tensor().

Parameters:

• dims – shape

• fill_value – scalar to fill the tensor

• dtype

• device

• sparse_dim

• feature_dim

returnn.frontend.const.fill(*, dims: Sequence[Dim], fill_value: int | float | complex | number | ndarray | bool | str | Tensor, dtype: str | None = None, device: str | None = None, sparse_dim: Dim | None = None, feature_dim: Dim | None = None) → Tensor

full, fill, constant.

https://data-apis.org/array-api/latest/API_specification/generated/array_api.full.html

Also see convert_to_tensor().

Parameters:

• dims – shape

• fill_value – scalar to fill the tensor

• dtype

• device

• sparse_dim

• feature_dim

returnn.frontend.const.constant(fill_value: int | float | complex | number | ndarray | bool | str, *, dims: Sequence[Dim], dtype: str | None = None, device: str | None = None, sparse_dim: Dim | None = None, feature_dim: Dim | None = None) → Tensor

alias to full(), mapping value to fill_value. also see convert_to_tensor()

returnn.frontend.const.zeros(dims: Sequence[Dim], *, dtype: str | None = None, device: str | None = None, sparse_dim: Dim | None = None, feature_dim: Dim | None = None) → Tensor

zeros. float by default.

returnn.frontend.const.ones(dims: Sequence[Dim], *, dtype: str | None = None, device: str | None = None, sparse_dim: Dim | None = None, feature_dim: Dim | None = None) → Tensor

ones. float by default.

returnn.frontend.const.zeros_like(other: Tensor) → Tensor

zeros like other

returnn.frontend.const.ones_like(other: Tensor) → Tensor

ones like other