| Safe Haskell | None |
|---|---|
| Language | Haskell2010 |
TensorFlow.Convolution
Synopsis
- data Padding
- data DataFormat
- conv2D :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 t -> Tensor Build t
- conv2D' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 t -> Tensor Build t
- conv2DBackpropFilter :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- conv2DBackpropFilter' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- conv2DBackpropInput :: OneOf '[Word16, Double, Float] t => Tensor v1 Int32 -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv2DBackpropInput' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 Int32 -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3D :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 t -> Tensor Build t
- conv3D' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 t -> Tensor Build t
- conv3DBackpropFilter :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3DBackpropFilter' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3DBackpropFilterV2 :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- conv3DBackpropFilterV2' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- conv3DBackpropInput :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3DBackpropInput' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3DBackpropInputV2 :: (OneOf '[Word16, Double, Float] t, OneOf '[Int32, Int64] tshape) => Tensor v1 tshape -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- conv3DBackpropInputV2' :: (OneOf '[Word16, Double, Float] t, OneOf '[Int32, Int64] tshape) => OpParams -> Padding -> DataFormat -> Tensor v1 tshape -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- depthwiseConv2dNative :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 t -> Tensor Build t
- depthwiseConv2dNative' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 t -> Tensor Build t
- depthwiseConv2dNativeBackpropFilter :: OneOf '[Word16, Double, Float] t => Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- depthwiseConv2dNativeBackpropFilter' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 t -> Tensor v2 Int32 -> Tensor v3 t -> Tensor Build t
- depthwiseConv2dNativeBackpropInput :: OneOf '[Word16, Double, Float] t => Tensor v1 Int32 -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
- depthwiseConv2dNativeBackpropInput' :: OneOf '[Word16, Double, Float] t => OpParams -> Padding -> DataFormat -> Tensor v1 Int32 -> Tensor v2 t -> Tensor v3 t -> Tensor Build t
Documentation
Convolution padding.
Constructors
| PaddingValid | output_spatial_shape[i] = ceil(input_spatial_shape[i] / strides[i]) |
| PaddingSame | output_spatial_shape[i] = ceil( (input_spatial_shape[i] - (spatial_filter_shape[i]-1) * dilation_rate[i]) / strides[i]) |
data DataFormat Source #
Matrix format.
Constructors
| ChannelLast | Channel is the last dimension (e.g. NWC, NHWC, NDHWC) |
| ChannelFirst | Channel is the first dimension after N (e.g. NCW, NCHW, NCDHW) |
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 t | filter |
| -> Tensor Build t | output |
2D Convolution with default parameters.
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 Int32 | filter_sizes |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
2D convolution backpropagation filter with default parameters.
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 Int32 | input_sizes |
| -> Tensor v2 t | filter |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
2D convolution backpropagation input with default parameters.
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 t | filter |
| -> Tensor Build t | output |
3D Convolution with default parameters.
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 t | filter |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
3D convolution backpropagation filter with default parameters.
conv3DBackpropFilterV2 Source #
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 Int32 | filter_sizes |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
3D convolution backpropagation filter with default parameters.
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 t | filter |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
3D convolution backpropagation input with default parameters.
conv3DBackpropInputV2 Source #
Arguments
| :: (OneOf '[Word16, Double, Float] t, OneOf '[Int32, Int64] tshape) | |
| => Tensor v1 tshape | input_sizes |
| -> Tensor v2 t | filter |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
3D convolution backpropagation input with default parameters.
depthwiseConv2dNative Source #
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 t | filter |
| -> Tensor Build t | output |
Depth-wise 2D convolution native with default parameters.
depthwiseConv2dNativeBackpropFilter Source #
Arguments
| :: OneOf '[Word16, Double, Float] t | |
| => Tensor v1 t | input |
| -> Tensor v2 Int32 | filter_sizes |
| -> Tensor v3 t | out_backprop |
| -> Tensor Build t | output |
Depth-wise 2D convolution native backpropagation filter with default parameters.