plaidml.op¶
Description
The TILE standard operation library.
These operations have been shown to be useful across a variety of frameworks. (Frameworks are of course free to define their own operations in addition to these, although it’ll be easier to use them with these if a framework’s own operations are defined using the standard plaidml.tile base classes.)
Each operation is defined as a tile.Operation subclass, allowing it to be
used in pattern matching. Additionally, each operation is provided via a
top-level function that wraps the class, allowing composite operations to
be built up using a functional programming style.
See the PlaidML Op Tutorial for information about writing your own custom operations.
Classes
ArgMax(value[, axis]) |
Maximum of elements along an axis. |
AutoPadding |
|
AveragePool(data, kernel_shape, pads, strides) |
A standard ML average pooling operator. |
BinaryCrossentropy(target, output, epsilon) |
Computes the binary crossentropy of a value relative to a target. |
Cast(x, dtype) |
|
ClipMax(value, max_val) |
Clips a Value to a maximum bound. |
ClipMin(value, min_val) |
Clips a Value to a minimum bound. |
Concatenate(tensors[, axis]) |
Concatenates tensors to make a single larger tensor. |
Convolution(data, kernel[, strides, …]) |
A standard ML convolution operator. |
ConvolutionDataFormat |
|
ConvolutionTranspose(x, kernel, …) |
A transposed convolution operator. |
CumulativeSum(x[, axis]) |
Cumulative sum of a tensor |
Dot(x, y) |
Dot-product of two tensors. |
Elu(x[, alpha]) |
Exponential linear unit. |
Equal(lhs, rhs) |
Elementwise tensor equality. |
Equal_ArgMax(lhs, rhs) |
|
Flatten(data) |
Flattens a tensor to a one-dimensional value. |
Gather(value, indicies) |
Gathers elements of a tensor. |
Gemm(a, b, c[, alpha, beta, broadcast, …]) |
Implements a general matrix multiplication. |
Gradients(loss, variables) |
Compute the gradients of a loss with respect to a set of values |
Hardmax(data) |
Implements a standard ML hardmax. |
Identity(x) |
A simple identity operation. |
IsMax(value, axes) |
True iff an input’s value is the maximum along some set of axes. |
LogSoftmax(data) |
Implements the log() of a standard ML softmax. |
MatMul(a, b) |
A matrix multiplication, using numpy semantics. |
MaxPool(data, padding, kernel_shape, pads, …) |
A standard ML max pooling operator. |
MaxReduce(x[, axes, keepdims]) |
Computes the maximum value along some set of axes. |
Mean(x[, axes, keepdims, floatx]) |
Computes the mean value along some set of axes. |
MinReduce(x[, axes, keepdims]) |
Computes the minimum value along some set of axes. |
NotEqual(lhs, rhs) |
Elementwise tensor inequality. |
Pow(x, p) |
An elementwise pow() function. |
Prod(value[, axes, keepdims, floatx]) |
|
Relu(x[, alpha, max_value]) |
A Rectified Linear Unit. |
Reshape(x, dims) |
Reshapes a tensor, without changing the type or number of elements. |
SliceTensor(data[, axes, ends, starts]) |
Implements tensor slicing. |
Softmax(data) |
Implements a standard ML softmax. |
Sqrt(x) |
Computes the elementwise square root of a value. |
Summation(value[, axes, keepdims, floatx]) |
Sums an input value along some set of axes. |
Variance(x[, axes, keepdims, floatx]) |
Functions
ceiling(data) |
Elementwise ceiling. |
clip(value, min_val, max_val) |
|
cos(data) |
Elementwise cosine. |
equal(lhs, rhs) |
Elementwise tensor equality. |
exp(data) |
Elementwise exponential. |
floor(data) |
Elementwise floor. |
gradients(loss, variables) |
|
hardmax(x[, axis]) |
|
log(data) |
Elementwise logarithm. |
log_softmax(x[, axis]) |
|
max_reduce(x[, axes, keepdims]) |
|
mean(x[, axes, keepdims, floatx]) |
|
min_reduce(x[, axes, keepdims]) |
|
pad_compute(sym, input_size, filter_size, …) |
Computes info for an axis of a padded filter. |
prod(value[, axes, keepdims, floatx]) |
|
sigmoid(data) |
Elementwise sigmoid. |
sin(data) |
Elementwise sine. |
softmax(x[, axis]) |
|
squeeze(x, axes) |
|
summation(value[, axes, keepdims, floatx]) |
|
tanh(data) |
Elementwise hyperbolic tangent. |
unsqueeze(x, axes) |