ngraph::pass namespace More...
Data Structures | |
class | AlgebraicSimplification |
class | BroadcastElementwiseFusion |
Removing Broadcast OP before ElementWise if output shape of Broadcast are equal neighboring input shape of ElementWise. More... | |
class | ClampFusion |
ClampFusion transformation replaces following graph: Maximum->Minimum to Clamp Restrictions: More... | |
class | CommonOptimizations |
class | ConvAddFusion |
class | ConvMultiplyFusion |
class | DeconvAddFusion |
class | ConvFusion |
class | ConvolutionMultiplyFusion |
class | GroupConvolutionMultiplyFusion |
class | ConvolutionBackpropDataMultiplyFusion |
class | GroupConvolutionBackpropDataMultiplyFusion |
class | ConvertQuantizeDequantize |
ConvertQuantizeDequantize transformation replaces following graph: FakeQuantize->Convert->Convert->Subtract->Multiply with a single FakeQuantize. Restrictions: More... | |
class | DepthToSpaceFusion |
DepthToSpaceFusion transformation detects Reshape-Transpose-Reshape pattern and tries to fuse it into a single DepthToSpace layer. More... | |
class | EliminateUnsqueezeGather |
Remove Unsqueeze + Gather pair, if Gather gathers data by dimension that was previously added by Unsqueeze. More... | |
class | FakeQuantizeMulFusion |
This transformation looks for a FQ + Mul pair in the graph and moves the Mul operation above the FQ node. The last two inputs of FQ are multiplied by the value that was originally below the FQ node. More... | |
class | FakeQuantizeReshapeFusion |
This transformation looks for a FQ + Reshape pair in the graph and moves the Reshape operation above the FQ node. Shapes of limit inputs are updated following FQ broadcasting semantics. More... | |
class | HSigmoidFusionWithReluDiv |
HSigmoidFusion transformation replaces a sub-graph (x * (min(Relu(x + 3), 6))) / 6 with a HSigmoid op. More... | |
class | HSigmoidFusionWithReluMul |
HSigmoidFusion transformation replaces a sub-graph (x * (min(Relu(x + 3), 6)) * const(1/6) with a HSigmoid op. More... | |
class | HSigmoidFusionWithoutRelu |
HSigmoidFusion transformation replaces a sub-graph x * (min(max(x + 3, 0), 6) / 6) with a HSigmoid op. More... | |
class | HSigmoidFusionWithClamp |
HSigmoidFusion transformation replaces a sub-graph x * (Clamp(x + 3, 0, 6) * const(1/6)) with a HSigmoid op. More... | |
class | HSigmoidFusion |
HSigmoidFusion transformation replaces various sub-graphs with a HSigmoid op. More... | |
class | HSwishFusionWithReluDiv |
HSwishFusion transformation replaces a sub-graph (x * (min(Relu(x + 3), 6))) / 6 with a HSwish op. More... | |
class | HSwishFusionWithReluMul |
HSwishFusion transformation replaces a sub-graph (x * (min(Relu(x + 3), 6)) * const(1/6) with a HSwish op. More... | |
class | HSwishFusionWithoutRelu |
HSwishFusion transformation replaces a sub-graph x * (min(max(x + 3, 0), 6) / 6) with a HSwish op. More... | |
class | HSwishFusionWithClampMul |
HSwishFusion transformation replaces a sub-graph x * (Clamp(x + 3, 0, 6) * const(1/6)) with a HSwish op. More... | |
class | HSwishFusionWithClampDiv |
HSwishFusion transformation replaces a sub-graph x * (Clamp(x + 3, 0, 6) / 6) with a HSwish op. More... | |
class | HSwishFusionWithHSigmoidMul |
HSwishFusion transformation replaces a sub-graph x * HSigmoid(x) with a HSwish op. More... | |
class | HSwishFusion |
HSwishFusion transformation replaces various sub-graphs with a HSwish op. More... | |
class | AddMultiplyFusion |
class | AddAddFusion |
class | MultiplyMultiplyFusion |
class | LinOpSequenceFusion |
class | MishFusion |
MishFusion transformation replaces group of operations: x * tanh(log(exp(x) + 1)) to Mish op. More... | |
class | NopElimination |
class | NormalizeL2FusionWithMax |
NormalizeL2FusionWithMax transformation replaces a sub-graph x/(max(sqrt(sum(x[j0, ..., jN]**2), eps)) with a NormalizeL2 op. More... | |
class | NormalizeL2FusionWithAdd |
NormalizeL2FusionWithAdd transformation replaces a sub-graph x/(add(sqrt(sum(x[j0, ..., jN]**2), eps)) with a NormalizeL2 op. More... | |
class | NormalizeL2Fusion |
NormalizeL2Fusion transformation replaces various sub-graphs with a NormalizeL2 op. More... | |
class | UselessStridedSliceEraser |
UselessStridedSliceEraser transformation removes StridedSlice operations with equal input and output shapes. More... | |
class | SharedStridedSliceEraser |
SharedStridedSliceEraser transformation replaces group of StridedSlice operations with first StridedSlice in this group. All SrtideSlices in this group must be equal and consume the same output port. More... | |
class | GroupedStridedSliceOptimizer |
GroupedStridedSliceOptimizer transformation replaces group of StridedSlice operations with VariadicSplit. All StridedSlice operations must slice data with the same axis and stride = 1. More... | |
class | StridedSliceOptimization |
StridedSliceOptimization transformation executes all transformations related to StridedSlice optimizations. More... | |
class | PadFusionAvgPool |
PadFusion transformation replaces following graph: Pad -> AvgPool to AvgPool, under following conditions. More... | |
class | PadFusionMaxPool |
PadFusion transformation replaces following graph: Pad -> MaxPool to MaxPool, under following conditions. More... | |
class | PadFusionConvolution |
PadFusion transformation replaces following graph: Pad -> Convolution to Convolution, under following conditions. More... | |
class | PadFusionConvolutionBackpropData |
PadFusion transformation replaces following graph: Pad -> ConvolutionBackpropData to ConvolutionBackpropData, under following conditions. More... | |
class | PadFusionGroupConvolution |
PadFusion transformation replaces following graph: Pad -> GroupConvolution to GroupConvolution, under following conditions. More... | |
class | PadFusionGroupConvolutionBackpropData |
PadFusion transformation replaces following graph: Pad -> GroupConvolutionBackpropData to GroupConvolutionBackpropData, under following conditions. More... | |
class | PadFusion |
class | PullTransposeThroughFQUp |
class | ReluFakeQuantizeFusion |
ReluFakeQuantizeFusion transformation replaces following graph: Relu -> FakeQuantize to FakeQuantize under following conditions: More... | |
class | RemoveFilteringBoxesBySize |
class | SoftPlusFusion |
SoftPlusFusion transformation replaces group of operations: log(exp(x) + 1) to SoftPlus op. More... | |
class | SoftPlusToMishFusion |
SoftPlusToMishFusion transformation replaces group of operations: x * tanh(softplus(x)) to Mish op. More... | |
class | SwishFusionWithSigmoid |
SwishFusionWithSigmoid replaces a sub-graphs x * Sigmoid(x) with a Swish op. More... | |
class | SwishFusionWithSigmoidWithBeta |
SwishFusionWithSigmoid replaces a sub-graphs x * Sigmoid(x * beta) with a Swish op. More... | |
class | SwishFusionWithBeta |
SwishFusionWithSigmoid replaces a sub-graphs x / (1.0 + exp(-x * beta)) with a Swish op. More... | |
class | SwishFusionWithoutBeta |
SwishFusionWithSigmoid replaces a sub-graphs x / (1.0 + exp(-x)) with a Swish op. More... | |
class | SwishFusion |
SwishFusion transformation replaces various sub-graphs with a Swish op. More... | |
class | WeightsDequantizeToFakeQuantize |
WeightsDequantizeToFakeQuantize transformation replaces Constant (i8) -> Convert (to fp) -> Subtract (zp) -> Multiply (scale) -> with Constant (i8) -> Convert (to fp) -> FakeQuantize -> deducing levels and FakeQuantize limits according to actual values in the weights Constant. More... | |
class | UnrollTensorIterator |
Unrolls the body of the TensorIterator layer. Multiple body copies, the number of which is determined by the number of iterations of the TensorIterator layer, are created and connected to each other and to the external network. If the number of TensorIterator iterations is greater than 1, then additional Concat and Split layers are added to the network. More... | |
class | ConvertPrecision |
class | InitNodeInfo |
InitNodeInfo transformation helps to set runtime info attributes in a single place. More... | |
class | DisableConvertConstantFoldingOnConstPath |
class | BatchNormDecomposition |
class | BidirectionalLSTMSequenceDecomposition |
Decompose LSTMSequence to forward and reverse LSTMSequence. More... | |
class | BidirectionalGRUSequenceDecomposition |
Decompose GRUSequence to forward and reverse GRUSequence. More... | |
class | BidirectionalRNNSequenceDecomposition |
Decompose RNNSequence to forward and reverse RNNSequence. More... | |
class | BidirectionalSequenceDecomposition |
Container for all types of sequences decomposition. More... | |
class | ConvertBatchToSpace |
ConvertBatchToSpace transformation decomposes BatchToSpace layer to Reshape->Transpose->Reshape->Crop. More... | |
class | ConvertBroadcast3 |
class | ConvertBroadcastToTiles |
class | ConvertConvolution |
class | ConvertGroupConvolution |
class | ConvertDeconvolution |
class | ConvertGroupDeconvolution |
class | ConvertConvolutions |
class | ConvertDepthToSpace |
class | ConvertDivide |
class | ConvertGather0D |
ConvertGather0D decomposes v1::Gather operation into v0::Unsqueeze + v1::Gather + v0::Squeeze pattern when gather indices is scalar. More... | |
class | ConvertGELU |
class | ConvertInterpolate1ToInterpolate4 |
ConvertInterpolate1ToInterpolate4 covert v0:interpolate into v4::Interpolate. More... | |
class | ConvertMinimum |
class | ConvertMod |
class | ConvertMVN1ToMVN6 |
ConvertMVN1ToMVN6 covert v0:MVN into v6::MVN. More... | |
class | ConvertNegative |
class | ConvertNMSToNMSIEInternal |
class | ConvertPadToGroupConvolution |
ConvertPadToGroupConvolution transformation replaces Pad operation with GroupConvolution but has some restrictions on Pad parameters: More... | |
class | ConvertNMS1ToNMS5 |
class | ConvertNMS3ToNMS5 |
class | ConvertNMS4ToNMS5 |
class | ConvertReduceMeanToPooling |
class | ConvertReduceMaxToPooling |
class | ConvertReduceSumToPooling |
class | ConvertReduceToPooling |
class | ConvertScatterElementsToScatter |
ConvertScatterElementsToScatter convert opset3::ScatterElementsUpdate to opset3::ScatterUpdate. More... | |
class | ConvertRNNSequenceToTensorIterator |
ConvertRNNSequenceToTensorIterator transformation converts RNNSequence layer to TensorIterator *. More... | |
class | ConvertGRUSequenceToTensorIterator |
ConvertGRUSequenceToTensorIterator transformation converts GRUSequence layer to TensorIterator *. More... | |
class | ConvertLSTMSequenceToTensorIterator |
ConvertLSTMSequenceToTensorIterator transformation converts LSTMSequence layer to TensorIterator *. More... | |
class | ConvertShapeOf3 |
class | ConvertShuffleChannels3 |
class | ConvertSpaceToBatch |
ConvertSpaceToBatch transformation decomposes SpaceToBatch layer to Pad->Reshape->Transpose->Reshape. More... | |
class | ConvertSpaceToDepth |
class | ConvertSubtract |
class | ConvertTensorIteratorToLSTMSequence |
Finds all TensorIterator layers, detects the pattern Squeeze->LSTMCell->Unsqueeze in the TensorIterator body, converts this pattern to LSTMSequence layer and replaces them TensorIterator. More... | |
class | ConvertTensorIteratorToRNNSequence |
Finds all TensorIterator layers, detects the pattern Squeeze->RNNCell->Unsqueeze in the TensorIterator body, converts this pattern to RNNSequence layer and replaces them TensorIterator. More... | |
class | ConvertTensorIteratorToGRUSequence |
Finds all TensorIterator layers, detects the pattern Squeeze->GRUCell->Unsqueeze in the TensorIterator body, converts this pattern to GRUSequence layer and replaces them TensorIterator. More... | |
class | ConvertTopK3 |
class | FakeQuantizeDecomposition |
FakeQuantizeDecomposition transformation decomposes FakeQuantize layer. More... | |
class | GRUCellDecomposition |
GRUCellDecomposition transformation decomposes GRUCell layer with inputs X, H, W, R, B to Add, Split, MatMul, Multiply and Subtract ops according to the formula: (.) - Denotes element-wise multiplication. More... | |
class | HSigmoidDecomposition |
HSigmoidDecomposition transformation into sub-graph (min(Relu(x + 3), 6) * const(1/6). More... | |
class | HSwishDecomposition |
HSwishDecomposition transformation into sub-graph x * (min(Relu(x + 3), 6) * const(1/6). More... | |
class | LogSoftmaxDecomposition |
LogSoftmaxDecomposition transformation into sub-graph x - log(reduce_sum(exp(x), axis)). More... | |
class | LSTMCellDecomposition |
LSTMCellDecomposition transformation decomposes LSTMCell layer with inputs X, H, C, W, R, B to Add, Split, MatMul, Multiply ops according to the formula: (.) - Denotes element-wise multiplication. More... | |
class | MVN6Decomposition |
MVN6Decomposition transformation into sub-graph x - ReduceMean(x, axes) if normalize_variance is false and into sub-graph (x - ReduceMean(x, axes)) / Sqrt(ReduceMean((x - ReduceMean(x, axes)) ^ 2)) if normalize_variance is true. More... | |
class | ReduceL1Decomposition |
Decomposes ReduceL1 into ReduceSum(abs(x)). More... | |
class | ReduceL2Decomposition |
Decomposes ReduceL2 into sqrt(ReduceSum(x * x)). More... | |
class | RNNCellDecomposition |
RNNCellDecomposition transformation decomposes RNNCell layer with inputs X, H, W, R, B to Add, MatMul ops according to the formula: More... | |
class | SimplifyCTCGreedyDecoderSeqLen |
SimplifyCTCGreedyDecoder converts v6:CTCGreedyDecoderSeqLen into v0::CTCGreedyDecoder. More... | |
class | SoftPlusDecomposition |
SoftPlusDecomposition transformation replaces SoftPlus op to group of operations: log(exp(x) + 1). More... | |
class | ConvertOpSet2ToOpSet1 |
class | ConvertOpSet3ToOpSet2 |
class | Serialize |
Serialize transformation converts ngraph::Function into IR files. More... | |
class | ReshapeAMatMul |
ReshapeAMatMul and ReshapeBMatMul transformations relax hard-coded Reshape followed by MatMul operation For 2D Reshape search patterns are: More... | |
class | ReshapeBMatMul |
class | TransposeMatMul |
class | MimicSetBatchSize |
MimicSetBatchSize transformation relaxes hard-coded output batch dimension of Reshape operation. For Reshape with input shape [in_batch, ...] and pattern value [out_batch, ...] it generates a sub-graph which basically keeps ratio of input and output batch size and performs the following calculation: More... | |
class | Proposal1Scales |
ProposalScales transformation helps to silently avoid reshape issues on the scale-input of Proposal layer. More... | |
class | Proposal4Scales |
class | ReshapeTo1D |
ReshapeTo1D transformation looks for Reshape from nD to 1D tensor and replaces its pattern to [-1]. More... | |
class | SetBatchSize |
Generic caller for all the transformations responsible to make model reshape-able by batch dimension. More... | |
class | SmartReshape |
class | StridedSliceSqueeze |
StridedSliceSqueeze transformation looks for SS -> Squeeze and corrects SS inputs and attributes for SS output to be squeeze-able. More... | |
class | SqueezeStridedSlice |
StridedSliceSqueeze transformation looks for Squeeze -> SSe and corrects SS inputs and attributes for SS output to be squeeze-able. More... | |
class | SharedSqueeze |
SharedSqueeze transformation looks for shared Squeezes and leaves only one Squeeze reconnecting all the outputs to it. More... | |
ngraph::pass namespace