[LPT] Documentation (developer guide)

docs/IE_PLUGIN_DG/ExecutableNetwork.md

@@ -37,7 +37,7 @@ The implementation `CompileNetwork` is fully device-specific.
 
 The function accepts a const shared pointer to `ngraph::Function` object and performs the following steps:
 
-1. Applies ngraph passes using `TransformNetwork` function, which defines plugin-specific conversion pipeline. 
+1. Applies ngraph passes using `TransformNetwork` function, which defines plugin-specific conversion pipeline. To support low precision inference, the pipeline can include Low Precision Transformations. These transformations are usually hardware specific. You can find how to use and configure Low Precisions Transformations in [Low Precision Transformations](@ref openvino_docs_IE_DG_lpt) guide.
 2. Maps the transformed graph to a backend specific graph representation (for example, to MKLDNN graph for Intel CPU).
 3. Allocates and fills memory for graph weights, backend specific memory handles and so on.
 

docs/IE_PLUGIN_DG/Intro.md

@@ -52,6 +52,7 @@ Detailed guides
 * [Build](@ref openvino_docs_ie_plugin_dg_plugin_build) a plugin library using CMake\*
 * Plugin and its components [testing](@ref openvino_docs_ie_plugin_dg_plugin_testing)
 * [Quantized networks](@ref openvino_docs_ie_plugin_dg_quantized_networks)
+* [Low precision transformations](@ref openvino_docs_IE_DG_lpt) guide
 * [Writing nGraph transformations](@ref ngraph_transformation) guide
 
 API References

docs/IE_PLUGIN_DG/layout.xml

@@ -4,7 +4,78 @@
         <tab type="usergroup" url="index.html" visibile="yes" title="GUIDE">
             <tab type="usergroup" url="index.html" title="Developer Guide for Inference Engine Plugin Library">
                 <tab type="user" url="@ref plugin" visibile="yes" title="Implement Plugin Functionality"/>
-                <tab type="user" url="@ref executable_network" visibile="yes" title="Implement Executable Network Functionality"/>
+                <tab type="user" url="@ref executable_network" visibile="yes" title="Implement Executable Network Functionality">
+                    <tab type="usergroup" title="Low Precision Transformations" url="@ref openvino_docs_IE_DG_lpt">
+                        <tab type="user" title="Attributes" url="@ref openvino_docs_IE_DG_lpt_attributes">
+                            <tab type="user" title="AvgPoolPrecisionPreserved" url="@ref openvino_docs_IE_DG_lpt_AvgPoolPrecisionPreserved"/>
+                            <tab type="user" title="IntervalsAlignment" url="@ref openvino_docs_IE_DG_lpt_IntervalsAlignment"/>
+                            <tab type="user" title="PerTensorQuantization" url="@ref openvino_docs_IE_DG_lpt_PerTensorQuantization"/>
+                            <tab type="user" title="PrecisionPreserved" url="@ref openvino_docs_IE_DG_lpt_PrecisionPreserved"/>
+                            <tab type="user" title="Precisions" url="@ref openvino_docs_IE_DG_lpt_Precisions"/>
+                            <tab type="user" title="QuantizationAlignment" url="@ref openvino_docs_IE_DG_lpt_QuantizationAlignment"/>
+                        </tab>
+                        <tab type="user" title="Step 1. Prerequisites transformations" url="@ref openvino_docs_IE_DG_lpt_step1_prerequisites">
+                            <tab type="user" title="LinOpSequenceFusion" url="@ref openvino_docs_IE_DG_lpt_LinOpSequenceFusion"/>
+                            <tab type="user" title="PullReshapeThroughDequantization" url="@ref openvino_docs_IE_DG_lpt_PullReshapeThroughDequantization"/>
+                            <tab type="user" title="PullTransposeThroughDequantization" url="@ref openvino_docs_IE_DG_lpt_PullTransposeThroughDequantization"/>
+                        </tab>
+                        <tab type="user" title="Step 2. Markup transformations" url="@ref openvino_docs_IE_DG_lpt_step2_markup">
+                            <tab type="user" title="AlignQuantizationIntervals" url="@ref openvino_docs_IE_DG_lpt_AlignQuantizationIntervals"/>
+                            <tab type="user" title="AlignQuantizationParameters" url="@ref openvino_docs_IE_DG_lpt_AlignQuantizationParameters"/>
+                            <tab type="user" title="CreateAttribute" url="@ref openvino_docs_IE_DG_lpt_CreateAttribute"/>
+                            <tab type="user" title="CreatePrecisionsDependentAttribute" url="@ref openvino_docs_IE_DG_lpt_CreatePrecisionsDependentAttribute"/>
+                            <tab type="user" title="MarkupAvgPoolPrecisionPreserved" url="@ref openvino_docs_IE_DG_lpt_MarkupAvgPoolPrecisionPreserved"/>
+                            <tab type="user" title="MarkupCanBeQuantized" url="@ref openvino_docs_IE_DG_lpt_MarkupCanBeQuantized"/>
+                            <tab type="user" title="MarkupPerTensorQuantization" url="@ref openvino_docs_IE_DG_lpt_MarkupPerTensorQuantization"/>
+                            <tab type="user" title="MarkupPrecisions" url="@ref openvino_docs_IE_DG_lpt_MarkupPrecisions"/>
+                            <tab type="user" title="PropagatePrecisions" url="@ref openvino_docs_IE_DG_lpt_PropagatePrecisions"/>
+                            <tab type="user" title="PropagateThroughPrecisionPreserved" url="@ref openvino_docs_IE_DG_lpt_PropagateThroughPrecisionPreserved"/>
+                            <tab type="user" title="PropagateToInput" url="@ref openvino_docs_IE_DG_lpt_PropagateToInput"/>
+                            <tab type="user" title="UpdateSharedPrecisionPreserved" url="@ref openvino_docs_IE_DG_lpt_UpdateSharedPrecisionPreserved"/>
+                        </tab>
+                        <tab type="user" title="Step 3. Main transformations" url="@ref openvino_docs_IE_DG_lpt_step3_main">
+                            <tab type="user" title="AddTransformation" url="@ref openvino_docs_IE_DG_lpt_AddTransformation"/>
+                            <tab type="user" title="AvgPoolTransformation" url="@ref openvino_docs_IE_DG_lpt_AvgPoolTransformation"/>
+                            <tab type="user" title="ClampTransformation" url="@ref openvino_docs_IE_DG_lpt_ClampTransformation"/>
+                            <tab type="user" title="ConcatTransformation" url="@ref openvino_docs_IE_DG_lpt_ConcatTransformation"/>
+                            <tab type="user" title="ConvolutionTransformation" url="@ref openvino_docs_IE_DG_lpt_ConvolutionTransformation"/>
+                            <tab type="user" title="ConvolutionBackpropDataTransformation" url="@ref openvino_docs_IE_DG_lpt_ConvolutionBackpropDataTransformation"/>
+                            <tab type="user" title="DepthToSpaceTransformation" url="@ref openvino_docs_IE_DG_lpt_DepthToSpaceTransformation"/>
+                            <tab type="user" title="FakeQuantizeDecompositionTransformation" url="@ref openvino_docs_IE_DG_lpt_FakeQuantizeDecompositionTransformation"/>
+                            <tab type="user" title="FakeQuantizeTransformation" url="@ref openvino_docs_IE_DG_lpt_FakeQuantizeTransformation"/>
+                            <tab type="user" title="InterpolateTransformation" url="@ref openvino_docs_IE_DG_lpt_InterpolateTransformation"/>
+                            <tab type="user" title="GroupConvolutionTransformation" url="@ref openvino_docs_IE_DG_lpt_GroupConvolutionTransformation"/>
+                            <tab type="user" title="MatMulTransformation" url="@ref openvino_docs_IE_DG_lpt_MatMulTransformation"/>
+                            <tab type="user" title="MaxPoolTransformation" url="@ref openvino_docs_IE_DG_lpt_MaxPoolTransformation"/>
+                            <tab type="user" title="MultiplyTransformation" url="@ref openvino_docs_IE_DG_lpt_MultiplyTransformation"/>
+                            <tab type="user" title="MVNTransformation" url="@ref openvino_docs_IE_DG_lpt_MVNTransformation"/>
+                            <tab type="user" title="NormalizeL2Transformation" url="@ref openvino_docs_IE_DG_lpt_NormalizeL2Transformation"/>
+                            <tab type="user" title="PadTransformation" url="@ref openvino_docs_IE_DG_lpt_PadTransformation"/>
+                            <tab type="user" title="PReluTransformation" url="@ref openvino_docs_IE_DG_lpt_PReluTransformation"/>
+                            <tab type="user" title="ReduceMaxTransformation" url="@ref openvino_docs_IE_DG_lpt_ReduceMaxTransformation"/>
+                            <tab type="user" title="ReduceMeanTransformation" url="@ref openvino_docs_IE_DG_lpt_ReduceMeanTransformation"/>
+                            <tab type="user" title="ReduceMinTransformation" url="@ref openvino_docs_IE_DG_lpt_ReduceMinTransformation"/>
+                            <tab type="user" title="ReduceSumTransformation" url="@ref openvino_docs_IE_DG_lpt_ReduceSumTransformation"/>
+                            <tab type="user" title="ReluTransformation" url="@ref openvino_docs_IE_DG_lpt_ReluTransformation"/>
+                            <tab type="user" title="ReshapeTransformation" url="@ref openvino_docs_IE_DG_lpt_ReshapeTransformation"/>
+                            <tab type="user" title="SqueezeTransformation" url="@ref openvino_docs_IE_DG_lpt_SqueezeTransformation"/>
+                            <tab type="user" title="ShuffleChannelsTransformation" url="@ref openvino_docs_IE_DG_lpt_ShuffleChannelsTransformation"/>
+                            <tab type="user" title="SplitTransformation" url="@ref openvino_docs_IE_DG_lpt_SplitTransformation"/>
+                            <tab type="user" title="StridedSliceTransformation" url="@ref openvino_docs_IE_DG_lpt_StridedSliceTransformation"/>
+                            <tab type="user" title="TransposeTransformation" url="@ref openvino_docs_IE_DG_lpt_TransposeTransformation"/>
+                            <tab type="user" title="UnsqueezeTransformation" url="@ref openvino_docs_IE_DG_lpt_UnsqueezeTransformation"/>
+                            <tab type="user" title="VariadicSplitTransformation" url="@ref openvino_docs_IE_DG_lpt_VariadicSplitTransformation"/>
+                        </tab>
+                        <tab type="user" title="Step 4. Cleanup transformations" url="@ref openvino_docs_IE_DG_lpt_step4_cleanup">
+                            <tab type="user" title="FoldConvertTransformation" url="@ref openvino_docs_IE_DG_lpt_FoldConvertTransformation"/>
+                            <tab type="user" title="FoldFakeQuantizeTransformation" url="@ref openvino_docs_IE_DG_lpt_FoldFakeQuantizeTransformation"/>
+                            <tab type="user" title="FuseConvertTransformation" url="@ref openvino_docs_IE_DG_lpt_FuseConvertTransformation"/>
+                            <tab type="user" title="FuseMultiplyToFakeQuantizeTransformation" url="@ref openvino_docs_IE_DG_lpt_FuseMultiplyToFakeQuantizeTransformation"/>
+                            <tab type="user" title="FuseSubtractToFakeQuantizeTransformation" url="@ref openvino_docs_IE_DG_lpt_FuseSubtractToFakeQuantizeTransformation"/>
+                            <tab type="user" title="MultiplyToGroupConvolutionTransformation" url="@ref openvino_docs_IE_DG_lpt_MultiplyToGroupConvolutionTransformation"/>
+                        </tab>
+                    </tab>
+                </tab>
                 <tab type="user" url="@ref infer_request" visibile="yes" title="Implement Synchronous Inference Request"/>
                 <tab type="user" url="@ref async_infer_request" visibile="yes" title="Implement Asynchronous Inference Request"/>
             </tab>

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/PluginTransformationPipeline.md

@@ -0,0 +1,17 @@
+# Plugin Transformation Pipeline {#openvino_docs_IE_DG_plugin_transformation_pipeline}
+
+@sphinxdirective
+
+.. toctree::
+   :maxdepth: 1
+   :caption: Executable Network
+   :hidden:
+
+   Low Precision Transformations <openvino_docs_IE_DG_lpt>
+
+@endsphinxdirective
+
+Typical plugin transformation pipeline includes steps:
+  1. Common transformations
+  2. [Low precision transformations](@ref openvino_docs_IE_DG_lpt)
+  3. Plugin specific transformations

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/avg_pool_precision_preserved.md

@@ -0,0 +1,11 @@
+# AvgPoolPrecisionPreserved attribute {#openvino_docs_IE_DG_lpt_AvgPoolPrecisionPreserved}
+
+ngraph::AvgPoolPrecisionPreservedAttribute class represents the `AvgPoolPrecisionPreserved` attribute.
+
+Utility attribute, which is used only during `AvgPool` operation, precision preserved property definition. 
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation                                    |
+| Properties    | value (boolean)                              |

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/intervals_alignment.md

@@ -0,0 +1,11 @@
+# IntervalsAlignment attribute {#openvino_docs_IE_DG_lpt_IntervalsAlignment}
+
+ngraph::IntervalsAlignmentAttribute class represents the `IntervalsAlignment` attribute.
+
+The attribute defines a subgraph with the same quantization intervals alignment. `FakeQuantize` operations are included. The attribute is used by quantization operations.
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation                                    |
+| Properties    | combined interval, minimal interval, minimal levels, preferable precisions |

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/per_tensor_quantization.md

@@ -0,0 +1,11 @@
+# PerTensorQuantization attribute {#openvino_docs_IE_DG_lpt_PerTensorQuantization}
+
+ngraph::PerTensorQuantizationAttribute class represents the `PerTensorQuantization` attribute.
+
+The attribute defines if the operation input port requires per-tensor quantization.
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation, input ports                        |
+| Properties    |                                              |

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/precision_preserved.md

@@ -0,0 +1,11 @@
+# PrecisionPreserved attribute {#openvino_docs_IE_DG_lpt_PrecisionPreserved}
+
+ngraph::PrecisionPreservedAttribute class represents the `PrecisionPreserved` attribute.
+
+The attribute defines a precision preserved operation. If the attribute is absent, then an operation is not precision preserved. 
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation                                    |
+| Properties    | value (boolean)                              |

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/precisions.md

@@ -0,0 +1,11 @@
+# Precisions attribute {#openvino_docs_IE_DG_lpt_Precisions}
+
+ngraph::PrecisionsAttribute class represents the `Precisions` attribute.
+
+The attribute defines precision which is required for input/output port or an operation.
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation, input port, output port           |
+| Properties    | precisions                                   |

docs/IE_PLUGIN_DG/plugin_transformation_pipeline/low_precision_transformations/attributes/quantization_alignment.md

@@ -0,0 +1,11 @@
+# QuantizationAlignment attribute {#openvino_docs_IE_DG_lpt_QuantizationAlignment}
+
+ngraph::QuantizationAlignmentAttribute class represents the `QuantizationAlignment` attribute.
+
+The attribute defines a subgraph with the same quantization alignment. `FakeQuantize` operations are not included. The attribute is used by quantization operations.
+
+| Property name | Values                                       |
+|---------------|----------------------------------------------|
+| Required      | Yes                                          |
+| Defined       | Operation                                    |
+| Properties    | value (boolean)                              |