Change KPI to Resource Utilization

README.md

@@ -117,7 +117,7 @@ This pruning technique is designed to compress models for specific hardware arch
 taking into account the target platform's Single Instruction, Multiple Data (SIMD) capabilities. 
 By pruning groups of channels (SIMD groups), our approach not only reduces model size 
 and complexity, but ensures that better utilization of channels is in line with the SIMD architecture 
-for a target KPI of weights memory footprint.
+for a target Resource Utilization of weights memory footprint.
 [Keras API](https://sony.github.io/model_optimization/docs/api/experimental_api_docs/methods/keras_pruning_experimental.html)
 [Pytorch API](https://github.com/sony/model_optimization/blob/main/model_compression_toolkit/pruning/pytorch/pruning_facade.py#L43) 
 

docsrc/source/api/api_docs/classes/KPI.rst → docsrc/source/api/api_docs/classes/ResourceUtilization.rst

@@ -1,11 +1,11 @@
 :orphan:
 
-.. _ug-KPI:
+.. _ug-ResourceUtilization:
 
 
-KPI
+ResourceUtilization
 ================================
 **Object to configure resources to use when searching for a configuration for the optimized model (such as in mixed-precision, pruning, etc.):**
 
-.. autoclass:: model_compression_toolkit.core.KPI
+.. autoclass:: model_compression_toolkit.core.ResourceUtilization
 

docsrc/source/api/api_docs/index.rst

@@ -49,12 +49,12 @@ core
 - :ref:`QuantizationConfig<ug-QuantizationConfig>`: Module to configure the quantization process.
 - :ref:`QuantizationErrorMethod<ug-QuantizationErrorMethod>`: Select a method for quantization parameters' selection.
 - :ref:`MixedPrecisionQuantizationConfig<ug-MixedPrecisionQuantizationConfig>`: Module to configure the quantization process when using mixed-precision PTQ.
-- :ref:`KPI<ug-KPI>`: Module to configure resources to use when searching for a configuration for the optimized model.
+- :ref:`ResourceUtilization<ug-ResourceUtilization>`: Module to configure resources to use when searching for a configuration for the optimized model.
 - :ref:`MpDistanceWeighting<ug-MpDistanceWeighting>`: Mixed precision distance metric weighting methods.
 - :ref:`network_editor<ug-network_editor>`: Module to modify the optimization process for troubleshooting.
 - :ref:`FolderImageLoader<ug-FolderImageLoader>`: Class to use an images directory as a representative dataset.
-- :ref:`pytorch_kpi_data<ug-pytorch_kpi_data>`: A function to compute KPI data that can be used to calculate the desired target KPI for PyTorch models.
-- :ref:`keras_kpi_data<ug-keras_kpi_data>`: A function to compute KPI data that can be used to calculate the desired target KPI for Keras models.
+- :ref:`pytorch_resource_utilization_data<ug-pytorch_resource_utilization_data>`: A function to compute Resource Utilization data that can be used to calculate the desired target resource utilization for PyTorch models.
+- :ref:`keras_resource_utilization_data<ug-keras_resource_utilization_data>`: A function to compute Resource Utilization data that can be used to calculate the desired target resource utilization for Keras models.
 
 
 data_generation

docsrc/source/api/api_docs/methods/keras_kpi_data.rst

@@ -1,10 +1,10 @@
 :orphan:
 
-.. _ug-keras_kpi_data:
+.. _ug-keras_resource_utilization_data:
 
 
-=======================================
-Get KPI information for Keras Models
-=======================================
+=======================================================
+Get Resource Utilization information for Keras Models
+=======================================================
 
-.. autofunction:: model_compression_toolkit.core.keras_kpi_data
+.. autofunction:: model_compression_toolkit.core.keras_resource_utilization_data

docsrc/source/api/api_docs/methods/pytorch_kpi_data.rst

@@ -1,10 +1,10 @@
 :orphan:
 
-.. _ug-pytorch_kpi_data:
+.. _ug-pytorch_resource_utilization_data:
 
 
-=======================================
-Get KPI information for PyTorch Models
-=======================================
+============================================================
+Get Resource Utilization information for PyTorch Models
+============================================================
 
-.. autofunction:: model_compression_toolkit.core.pytorch_kpi_data
+.. autofunction:: model_compression_toolkit.core.pytorch_resource_utilization_data

model_compression_toolkit/constants.py

@@ -93,7 +93,7 @@
 DEC_RANGE_BOTTOM = 0.97
 DEC_RANGE_UPPER = 1.03
 
-# KPI computation parameters
+# Resource utilization computation parameters
 BITS_TO_BYTES = 8.0
 
 # Default threshold for Softmax layer

model_compression_toolkit/core/__init__.py

@@ -21,9 +21,9 @@
 from model_compression_toolkit.core.common.mixed_precision import mixed_precision_quantization_config
 from model_compression_toolkit.core.common.quantization.quantization_config import QuantizationConfig, QuantizationErrorMethod, DEFAULTCONFIG
 from model_compression_toolkit.core.common.quantization.core_config import CoreConfig
-from model_compression_toolkit.core.common.mixed_precision.kpi_tools.kpi import KPI
+from model_compression_toolkit.core.common.mixed_precision.resource_utilization_tools.resource_utilization import ResourceUtilization
 from model_compression_toolkit.core.common.mixed_precision.mixed_precision_quantization_config import MixedPrecisionQuantizationConfig
-from model_compression_toolkit.core.keras.kpi_data_facade import keras_kpi_data
-from model_compression_toolkit.core.pytorch.kpi_data_facade import pytorch_kpi_data
+from model_compression_toolkit.core.keras.resource_utilization_data_facade import keras_resource_utilization_data
+from model_compression_toolkit.core.pytorch.resource_utilization_data_facade import pytorch_resource_utilization_data
 from model_compression_toolkit.core.common.mixed_precision.distance_weighting import MpDistanceWeighting
 

model_compression_toolkit/core/common/graph/virtual_activation_weights_node.py

@@ -113,14 +113,14 @@ def __init__(self, origin_node: BaseNode, activation_class: type, fw_attr: dict)
 class VirtualActivationWeightsNode(BaseNode):
     """
     A node that represents a composition of pair of sequential activation node and weights (kernel) node.
-    This structure is used for mixed-precision search with bit-operation KPI.
+    This structure is used for mixed-precision search with bit-operation constraint.
     The node's candidates are the cartesian product of both nodes' candidates.
 
     Important: note that not like regular BaseNode or FunctionalNode, in VirtualActivationWeightsNode the activation
     candidates config refer to the quantization config of the activation that precedes the linear operation! instead of
     the output of the linear operation.
     It is ok, since this node is not meant to be used in a graph for creating an actual model, but only a virtual
-    representation of the model's graph only for allowing to compute the bit-operations KPI in mixed-precision.
+    representation of the model's graph only for allowing to compute the bit-operations constraint in mixed-precision.
     """
 
     def __init__(self,