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

Python package

max.tensor

Distributed tensors and the eager runtime that realizes them.

Tensor is the user-facing value type. Ops on tensors live in max.functional; placements and the device mesh live in max.tensor.sharding; ahead-of-time compilation of tensor programs lives in max.tensor.compilation.

The following example calls Tensor.ones() to build a 4x8 tensor and passes it to F.matmul along with its transpose:

from max import functional as F
from max.tensor import Tensor

x = Tensor.ones((4, 8))
y = F.matmul(x, x.T)

Submodules

`max.tensor.compilation`	Traces and compiles Python functions over tensors.
`max.tensor.executor`	Executor protocol and concrete implementations for eager graph execution.
`max.tensor.realization_context`	Implementations of various realization strategies.
`max.tensor.sharding`	Defines how a tensor is laid out across a device mesh and how ops reshard it.
`max.tensor.sharding.per_shard_dim`	Wraps one `Dim` per device, so a dimension can vary by shard.
`max.tensor.testing`	Testing utilities for the `Tensor` API.
`max.tensor.validation`	Guard rails for running eager code in production.

`Tensor`	A multi-dimensional array of numeric values on a CPU or accelerator device.

`default_device`	Context manager for setting the default device for tensor creation.
`default_dtype`	Context manager for setting the default dtype for tensor creation.
`defaults`	Gets the default dtype and device for tensor creation.
`defaults_like`	Context manager setting the default dtype and device for tensor creation.