torchtraining.cast

Cast tensors in a functional fashion.

Users can use this module to cast step outputs to desired type or to lower precision in order to save memory (though it shouldn’t be needed.)

class torchtraining.cast.As(other)

Cast torch.Tensor to the same type as other.

Parameters

other (torch.Tensor) – Tensor according to which incoming tensor will be casted.

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.BFloat16(memory_format=torch.preserve_format)

Cast torch.Tensor instance to BFloat16.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

BFloat16

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Bool(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Bool.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Bool

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Byte(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Byte.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Byte

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Char(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Char.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Char

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Double(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Double.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Double

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Float(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Float.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Float

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

torchtraining.cast.Float16

alias of torchtraining.cast.Half

torchtraining.cast.Float32

alias of torchtraining.cast.Float

torchtraining.cast.Float64

alias of torchtraining.cast.Double

class torchtraining.cast.Half(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Half.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Half

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Int(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Int.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Int

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

torchtraining.cast.Int16

alias of torchtraining.cast.Short

torchtraining.cast.Int32

alias of torchtraining.cast.Int

torchtraining.cast.Int64

alias of torchtraining.cast.Long

torchtraining.cast.Int8

alias of torchtraining.cast.Char

class torchtraining.cast.Item

Cast torch.Tensor instance to Item.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Item

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.List

Cast torch.Tensor instance to List.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

List

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Long(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Long.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Long

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.MKLDNN

Cast torch.Tensor instance to MKLDNN.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

MKLDNN

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Numpy

Cast torch.Tensor instance to Numpy.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Numpy

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Short(memory_format=torch.preserve_format)

Cast torch.Tensor instance to Short.

Note

IMPORTANT: Only torch.Tensor can be passed as memory_format is specified during casting.

Returns

Casted data

Return type

Short

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

class torchtraining.cast.Sparse(sparse_dims=None)

Cast torch.Tensor to sparse format.

Parameters

sparse_dims (int, optional) – The number of sparse dimensions to include in the new sparse tensor. Default: None.

forward(data)

Parameters

data (torch.Tensor) – Tensor to be casted

torchtraining.cast.UInt8

alias of torchtraining.cast.Byte