# tensorplay.distributed.tensor API

Source: https://www.tensorplay.cn/docs/api/tensorplay.distributed.tensor.html

## Functions 11

[#](#api-tensorplay.distributed.tensor.distribute_module)

### distribute_module

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.distribute_module.html)

```python
tensorplay.distributed.tensor.distribute_module(module: Any, device_mesh: DeviceMesh | None = None, partition_fn: Callable[[str, Any, DeviceMesh], Any] | None = None, input_fn: Callable[[Any, Any, DeviceMesh], Any] | None = None, output_fn: Callable[[Any, Any, DeviceMesh], Any] | None = None) → Any
```

[#](#api-tensorplay.distributed.tensor.distribute_tensor)

### distribute_tensor

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.distribute_tensor.html)

```python
tensorplay.distributed.tensor.distribute_tensor(tensor: Any, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None, *, src_data_rank: int | None = 0) → DTensor
```

[#](#api-tensorplay.distributed.tensor.empty)

### empty

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.empty.html)

```python
tensorplay.distributed.tensor.empty(*size: Any, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.from_local)

### from_local

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.from_local.html)

```python
tensorplay.distributed.tensor.from_local(local_tensor: Any, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None, *, run_check: bool = False, shape: Sequence[int] | None = None, stride: Sequence[int] | None = None, grad_placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.full)

### full

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.full.html)

```python
tensorplay.distributed.tensor.full(size: Any, fill_value: Any, *, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.linspace)

### linspace

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.linspace.html)

```python
tensorplay.distributed.tensor.linspace(start: Any, end: Any, steps: int, *, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.logspace)

### logspace

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.logspace.html)

```python
tensorplay.distributed.tensor.logspace(start: Any, end: Any, steps: int, base: float = 10.0, *, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.ones)

### ones

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.ones.html)

```python
tensorplay.distributed.tensor.ones(*size: Any, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.rand)

### rand

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.rand.html)

```python
tensorplay.distributed.tensor.rand(*size: Any, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.randn)

### randn

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.randn.html)

```python
tensorplay.distributed.tensor.randn(*size: Any, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

[#](#api-tensorplay.distributed.tensor.zeros)

### zeros

function[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.zeros.html)

```python
tensorplay.distributed.tensor.zeros(*size: Any, dtype: Any = None, layout: Any = 'strided', requires_grad: bool = False, device_mesh: DeviceMesh | None = None, placements: Sequence[Placement] | None = None) → DTensor
```

## Classes 5

[#](#api-tensorplay.distributed.tensor.DTensor)

### DTensor

class[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.DTensor.html)

```python
class tensorplay.distributed.tensor.DTensor(local_tensor: Any, device_mesh: DeviceMesh, placements: Sequence[Placement], *, shape: Sequence[int] | None = None, stride: Sequence[int] | None = None, grad_placements: Sequence[Placement] | None = None, backward_dtype: Any = None)
```

A logical tensor represented by a local value and a mesh placement.

[#](#api-tensorplay.distributed.tensor.Partial)

### Partial

class[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.Partial.html)

```python
class tensorplay.distributed.tensor.Partial(reduce_op: str = 'sum')
```

Store values that still need a reduction across one mesh dimension.

[#](#api-tensorplay.distributed.tensor.Placement)

### Placement

class[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.Placement.html)

```python
class tensorplay.distributed.tensor.Placement
```

Base class for a tensor layout on one mesh dimension.

[#](#api-tensorplay.distributed.tensor.Replicate)

### Replicate

class[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.Replicate.html)

```python
class tensorplay.distributed.tensor.Replicate
```

Keep a complete copy of the logical tensor on every rank.

[#](#api-tensorplay.distributed.tensor.Shard)

### Shard

class[Full reference ↗](/docs/generated/tensorplay.distributed.tensor.Shard.html)

```python
class tensorplay.distributed.tensor.Shard(dim: int)
```

Split one logical tensor dimension across a mesh dimension.