pymoose package

Subpackages

• pymoose.computation package
  • Submodules
  • Module contents
• pymoose.edsl package
  • Submodules
  • Module contents
• pymoose.predictors package
  • Submodules
  • Module contents
• pymoose.logger module
• pymoose.runtime module

eDSL Values & Types

Argument(placement[, dtype, vtype])	A type annotation Arguments to Moose computations.
AesKeyType()	Moose value representing an AES key.
AesTensorType(dtype)	Moose Value representing a tensor of AES-encrypted values.
FloatType()	Moose value representing a Python float.
IntType()	Moose value representing a Python integer.
TensorType(dtype)	Moose Value representing a Tensor.
StringType()	Moose value representing a Python string.

eDSL DTypes

bool_	Generic implementation of a Moose DType
fixed(integ, frac)	Factory function for creating a fixedpoint DType.
float32	Generic implementation of a Moose DType
float64	Generic implementation of a Moose DType
int32	Generic implementation of a Moose DType
int64	Generic implementation of a Moose DType
uint64	Generic implementation of a Moose DType
ring64	Generic implementation of a Moose DType

eDSL Placement Factories

host_placement(name)
mirrored_placement(name, players)
replicated_placement(name, players)

eDSL Functions

abs(x[, placement])	Compute the absolute value.
add(lhs, rhs[, placement])	Add two values.
add_n(arrays[, placement])	Elementwise addition of a collection of tensors.
argmax(x, axis, upmost_index[, placement])	Compute the index of the maximal element of a tensor along a given dimension.
atleast_2d(x[, to_column_vector, placement])
cast(x, dtype[, placement])	Cast a tensor to a new dtype.
concatenate(arrays[, axis, placement])	Concatenation of a collection of tensors along a given axis/dimension.
constant(value[, dtype, vtype, placement])	Embed a constant of a particular type.
decrypt(key, ciphertext[, placement])
div(lhs, rhs[, placement])	Divide two tensors.
dot(lhs, rhs[, placement])	Dot product of two tensors.
exp(x[, placement])	Elementiwise exponential function: \(e^x\).
expand_dims(x, axis[, placement])	Expand the rank of x with new singleton dimensions.
greater(lhs, rhs[, placement])	Evaluate the boolean greater-than operation, i.e. lhs > rhs.
identity(x[, placement])	The identity operation, f(x) = x.
index_axis(x, axis, index[, placement])	Index a tensor along a given dimension.
inverse(x[, placement])	Invert a floating-point matrix.
less(lhs, rhs[, placement])	Evaluate the boolean less-than operation, i.e. lhs < rhs.
load(key[, query, dtype, vtype, placement])	Load a value from placement storage.
log(x[, placement])	Compute the elementwise natural logarithm of a tensor.
log2(x[, placement])	Compute the elementwise base-2 logarithm of a tensor.
logical_and(lhs, rhs[, placement])	Evaluate the boolean AND operation, i.e. lhs & rhs.
logical_or(lhs, rhs[, placement])	Evaluate the boolean OR operation, i.e. lhs | rhs.
maximum(arrays[, placement])	Elementwise maximum of a collection of input tensors.
mean(x[, axis, placement])	Mean-reduce an input tensor.
mul(lhs, rhs[, placement])	Multiply two values.
mux(selector, x, y[, placement])	Multiplex two tensors according to some condition.
ones(shape, dtype[, placement])	Embed a ones array into the Moose computation graph.
output(tag, value[, placement])	Tag an output of a computation.
relu(x[, placement])	Apply the rectified-linear unit (ReLU) function, f(x) = max(0, x).
reshape(x, shape[, placement])	Reshape a tensor.
save(key, value[, placement])	Save a key-value pair to placement storage.
select(x, axis, index[, placement])	Select elements along some axis of a tensor according to some index tensor.
shape(x[, placement])	Compute the shape of an input tensor.
sliced(x, begin, end[, placement])	Compute a slice of a value.
softmax(x, axis, upmost_index[, placement])	Softmax function.
square(x[, placement])	Square an input value.
squeeze(x[, axis, placement])	Reduce out any singleton dimensions of x.
strided_slice(x, slices[, placement])	Compute a strided slice of a value.
sigmoid(x[, placement])	Apply the sigmoid function, \(\frac{1}{1 + e^{-x}}\).
sub(lhs, rhs[, placement])	Subtract two values.
sum(x[, axis, placement])	Sum-reduce an input tensor.
sqrt(x[, placement])	Compute the square-root of a value \(\sqrt{x}\).
transpose(x[, placement])	Compute the transpose of a tensor.
zeros(shape, dtype[, placement])	Embed a zeros array into the Moose computation graph.

PyMoose Context

get_current_placement()
get_current_runtime()	Get a global runtime context.
set_current_runtime(runtime)	Set a global runtime context.

Moose Bindings

computation([func, role_map])	Annotates a Python function as a Moose computation.
elk_compiler
GrpcMooseRuntime(identities)	Moose runtime backed by gRPC choreography.
LocalMooseRuntime(identities[, storage_mapping])	Locally-simulated Moose runtime.
MooseComputation