Go to the documentation of this file.
26 #ifndef MXNET_COMMON_ALM_H_
27 #define MXNET_COMMON_ALM_H_
34 #include <unordered_map>
88 std::vector<Transpose>* in_axes,
89 std::vector<Transpose>* out_axes)>;
100 #endif // MXNET_COMMON_ALM_H_
namespace of mxnet
Definition: api_registry.h:33
Transpose FactorCommonTranspose(std::vector< Transpose > *axes)
Factors out and returns a common transpose, or default-constructed Transpose if all axes (in/out para...
mshadow::LayoutFlag ApplyTranspose(mshadow::LayoutFlag layout, const Transpose &axes)
std::function< bool(nnvm::NodeAttrs *, mshadow::LayoutFlag target_layout, std::vector< Transpose > *in_axes, std::vector< Transpose > *out_axes)> FChangeLayout
May change operator's layout. Used in LayoutOptimization.
Definition: alm.h:89
Symbolic computation graph. This is the intermediate representation for optimization pass.
Definition: graph.h:47
nnvm::Graph OptimizeLayout(nnvm::Graph &&g)
LayoutFlag
Definition: base.h:498
Transpose FromTShape(const mxnet::TShape &s)
A singleton flag, set and read by MXSetOptimizeLayout and MXGetOptimizeLayout.
Definition: alm.h:43
The attributes of the current operation node. Usually are additional parameters like axis,...
Definition: node.h:107
Transpose Reverse(const Transpose &axes)
std::vector< size_t > Transpose
Transpose, represented by permutation of axes.
Definition: alm.h:60
Transpose Compose(const Transpose &lhs, const Transpose &rhs)
Configuation of nnvm as well as basic data structure.
A Shape class that is used to represent shape of each tensor.
Definition: tuple.h:440
bool optimize
Definition: alm.h:44
Graph node data structure.
bool IsIdentity(const Transpose &t)
configuration of MXNet as well as basic data structure.
static ALMParams & get()
Definition: alm.h:46