triton.experimental.gluon.nvidia.blackwell.TensorDescriptor

class triton.experimental.gluon.nvidia.blackwell.TensorDescriptor(base: Any, shape: List[int], strides: List[int], block_shape: List[int], layout: triton.experimental.gluon.language._layouts.NVMMASharedLayout, padding: str = 'zero', round_f32_to_tf32: bool = False)

__init__(self, base: Any, shape: List[int], strides: List[int], block_shape: List[int], layout: NVMMASharedLayout, padding: str = 'zero', round_f32_to_tf32: bool = False) → None

Methods

`__init__`(self, base, shape, strides, ...[, ...])
`from_tensor`(tensor, block_shape, layout[, ...])	Create a TensorDescriptor from a tensor.

Attributes

`mode`
`padding`
`round_f32_to_tf32`
`base`
`shape`
`strides`
`block_shape`
`layout`