For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /max/get-started.md).

Python class

DistributedTensorType

DistributedTensorType

class max.experimental.sharding.DistributedTensorType(dtype, shape, mesh, placements)

source

Bases: DistributedType[TensorType]

A symbolic type for a tensor distributed across a device mesh.

Analogous to TensorType for single-device tensors. Derives per-device TensorType objects via local_types.

When a SymbolicDim is sharded along a mesh axis, the local shard dimension becomes a new SymbolicDim named "{original}_{axis_name}". This keeps symbolic names short and debuggable while ensuring that sharding the same global dim on different axes produces distinct names.

Parameters:

• dtype (DType) – Element data type.
• shape (Shape) – Global (logical) shape.
• mesh (DeviceMesh) – The device mesh.
• placements (tuple[Placement, ...]) – One Placement per mesh axis.

local_types

property local_types: list[TensorType]

source

The per-device TensorType objects in mesh order.

Was this page helpful?

Thank you! We'll create more content like this.

Thank you for helping us improve!

😔 What went wrong?

Some code doesn’t work

It includes inaccurate information

It's missing information I need

It was difficult to understand

Other

Submit