# tensorplay._shape_funcs API

Source: https://www.tensorplay.cn/docs/api/tensorplay._shape_funcs.html

## Functions 5

[#](#api-tensorplay._shape_funcs.block_diag)

### block_diag

function[Full reference ↗](/docs/generated/tensorplay._shape_funcs.block_diag.html)

```python
tensorplay._shape_funcs.block_diag(*tensors)
```

Builds a block diagonal matrix from the given blocks.

0-D blocks become 1x1 matrices and 1-D blocks become single rows.  The
blocks are promoted to one type and must share a device; the zeros around
them are made on that device by the operator itself.

[#](#api-tensorplay._shape_funcs.broadcast_shapes)

### broadcast_shapes

function[Full reference ↗](/docs/generated/tensorplay._shape_funcs.broadcast_shapes.html)

```python
tensorplay._shape_funcs.broadcast_shapes(*shapes)
```

Returns the broadcast shape of the given shapes (right-aligned).

[#](#api-tensorplay._shape_funcs.broadcast_tensors)

### broadcast_tensors

function[Full reference ↗](/docs/generated/tensorplay._shape_funcs.broadcast_tensors.html)

```python
tensorplay._shape_funcs.broadcast_tensors(*tensors)
```

Broadcasts the given tensors to a common shape.

[#](#api-tensorplay._shape_funcs.tensordot)

### tensordot

function[Full reference ↗](/docs/generated/tensorplay._shape_funcs.tensordot.html)

```python
tensorplay._shape_funcs.tensordot(input, other, dims=2)
```

Contracts input and other over the given dimensions.

dims may be a non-negative int (contract the last dims dims of
input with the first dims dims of other) or a pair
(dims_a, dims_b) of dimension lists paired positionally.

[#](#api-tensorplay._shape_funcs.unravel_index)

### unravel_index

function[Full reference ↗](/docs/generated/tensorplay._shape_funcs.unravel_index.html)

```python
tensorplay._shape_funcs.unravel_index(indices, shape)
```

Converts flat indices into coordinate tuples (one LongTensor per dim).