Deep.Net


LoopEval

Namespace: SymTensor

functions for loop evaluation

Nested types and modules

TypeDescription
LoopChannelInfoT

channel information for loop execution

LoopChannelLayoutInfoT

channel layout information for building loop strides

Functions and values

Function or valueDescription
buildInOut (...)
Signature: nIters:int64 -> iter:int64 -> iterAry:ITensor -> itersRemainingAry:ITensor -> vars:Map<VarSpecT,LoopInputT> -> args:ITensor list -> channels:Map<ChannelT,LoopChannelInfoT> -> VarEnvT * Map<ChannelT,ITensor>

builds inputs and outputs for one loop iteration

buildStrides vars args channels
Signature: vars:Map<VarSpecT,LoopInputT> -> args:TensorLayout list -> channels:Map<ChannelT,LoopChannelLayoutInfoT> -> VarStridesT * ChannelStridesT * int list option list

build strides information for loop sources and targets

Fork me on GitHub