mct-nightly 1.8.0.8032023.post421__py3-none-any.whl → 1.8.0.8052023.post414__py3-none-any.whl

model_compression_toolkit/qat/pytorch/quantizer/ste_rounding/uniform_ste.py

@@ -12,18 +12,18 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from typing import Dict
-
 import numpy as np
 import torch
 import torch.nn as nn
 from torch import Tensor
 
-from model_compression_toolkit.core.common.constants import RANGE_MAX, RANGE_MIN
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
-from model_compression_toolkit.qat.common.constants import FQ_MIN, FQ_MAX
-from model_compression_toolkit.core.common import constants as C
-from model_compression_toolkit import quantizers_infrastructure as qi, TrainingMethod
+from model_compression_toolkit.constants import RANGE_MAX, RANGE_MIN
+from model_compression_toolkit.quantizers_infrastructure.constants import FQ_MIN, FQ_MAX
+
+from model_compression_toolkit.qat import TrainingMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
+from model_compression_toolkit import quantizers_infrastructure as qi, constants as C
+
 from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_quantizer import BasePytorchQATTrainableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 from model_compression_toolkit.core.pytorch.utils import to_torch_tensor
@@ -32,6 +32,7 @@ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructur
     WeightsUniformInferableQuantizer, ActivationUniformInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.trainable_infrastructure.common.trainable_quantizer_config import \
     TrainableQuantizerWeightsConfig, TrainableQuantizerActivationConfig
+from model_compression_toolkit.quantizers_infrastructure.trainable_infrastructure.common.base_trainable_quantizer import VariableGroup
 
 
 @mark_quantizer(quantization_target=qi.QuantizationTarget.Weights,
@@ -64,22 +65,18 @@ class STEUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
                               [-1]) if self.quantization_config.weights_per_channel_threshold else float(
             self.min_values)
 
-        self.quantizer_parameters = {}
 
     def initialize_quantization(self,
                                 tensor_shape: torch.Size,
                                 name: str,
-                                layer: qi.PytorchQuantizationWrapper) -> Dict[str, nn.Parameter]:
+                                layer: qi.PytorchQuantizationWrapper):
         """
-        Add min and max variables to layer.
-        Args:
-            tensor_shape: Tensor shape the quantizer quantize.
-            name: Prefix of variables names.
-            layer: Layer to add the variables to. The variables are saved
-            in the layer's scope.
+        Add quantizer parameters to the quantizer parameters dictionary
 
-        Returns:
-            Dictionary of new variables.
+        Args:
+            tensor_shape: tensor shape of the quantized tensor.
+            name: Tensor name.
+            layer: Layer to quantize.
         """
 
         # Add min and max variables to layer.
@@ -87,9 +84,9 @@ class STEUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
         layer.register_parameter(name+"_"+FQ_MAX, nn.Parameter(to_torch_tensor(self.max_values), requires_grad=False))
 
         # Save the quantizer parameters for later calculations
-        self.quantizer_parameters = {FQ_MIN: layer.get_parameter(name+"_"+FQ_MIN), FQ_MAX: layer.get_parameter(name+"_"+FQ_MAX)}
+        self.add_quantizer_variable(FQ_MIN, layer.get_parameter(name+"_"+FQ_MIN), VariableGroup.QPARAMS)
+        self.add_quantizer_variable(FQ_MAX, layer.get_parameter(name+"_"+FQ_MAX), VariableGroup.QPARAMS)
 
-        return self.quantizer_parameters
 
     def __call__(self,
                  inputs: nn.Parameter,
@@ -102,7 +99,7 @@ class STEUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
         Returns:
             quantized tensor
         """
-        return uniform_quantizer(inputs, self.quantizer_parameters[FQ_MIN], self.quantizer_parameters[FQ_MAX], self.num_bits)
+        return uniform_quantizer(inputs, self.get_quantizer_variable(FQ_MIN), self.get_quantizer_variable(FQ_MAX), self.num_bits)
 
     def convert2inferable(self) -> WeightsUniformInferableQuantizer:
         """
@@ -111,8 +108,8 @@ class STEUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
         Returns:
             A pytorch inferable quanizer object.
         """
-        _min = self.quantizer_parameters[FQ_MIN].cpu().detach().numpy()
-        _max = self.quantizer_parameters[FQ_MAX].cpu().detach().numpy()
+        _min = self.get_quantizer_variable(FQ_MIN).cpu().detach().numpy()
+        _max = self.get_quantizer_variable(FQ_MAX).cpu().detach().numpy()
 
         return WeightsUniformInferableQuantizer(num_bits=self.num_bits,
                                                 min_range=_min, max_range=_max,
@@ -143,21 +140,25 @@ class STEUniformActivationQATQuantizer(BasePytorchQATTrainableQuantizer):
         self.min_range_tensor = torch.Tensor([np_min_range])
         self.max_range_tensor = torch.Tensor([np_max_range])
         self.num_bits = quantization_config.activation_n_bits
-        self.quantizer_parameters = {}
 
     def initialize_quantization(self,
                                 tensor_shape: torch.Size,
                                 name: str,
-                                layer: qi.PytorchQuantizationWrapper) -> Dict[str, nn.Parameter]:
+                                layer: qi.PytorchQuantizationWrapper):
         """
-        Add min and max variables to layer.
+        Add quantizer parameters to the quantizer parameters dictionary
+
+        Args:
+            tensor_shape: tensor shape of the quantized tensor.
+            name: Tensor name.
+            layer: Layer to quantize.
         """
         layer.register_parameter(name+"_"+FQ_MIN, nn.Parameter(to_torch_tensor(self.min_range_tensor), requires_grad=True))
         layer.register_parameter(name+"_"+FQ_MAX, nn.Parameter(to_torch_tensor(self.max_range_tensor), requires_grad=True))
 
         # Save the quantizer parameters for later calculations
-        self.quantizer_parameters = {FQ_MIN: layer.get_parameter(name+"_"+FQ_MIN), FQ_MAX: layer.get_parameter(name+"_"+FQ_MAX)}
-        return self.quantizer_parameters
+        self.add_quantizer_variable(FQ_MIN, layer.get_parameter(name+"_"+FQ_MIN), VariableGroup.QPARAMS)
+        self.add_quantizer_variable(FQ_MAX, layer.get_parameter(name+"_"+FQ_MAX), VariableGroup.QPARAMS)
 
     def __call__(self,
                  inputs: torch.Tensor,
@@ -172,8 +173,8 @@ class STEUniformActivationQATQuantizer(BasePytorchQATTrainableQuantizer):
             The quantized tensor.
         """
 
-        _min = self.quantizer_parameters[FQ_MIN]
-        _max = self.quantizer_parameters[FQ_MAX]
+        _min = self.get_quantizer_variable(FQ_MIN)
+        _max = self.get_quantizer_variable(FQ_MAX)
         q_tensor = uniform_quantizer(inputs, _min, _max, self.num_bits)
         return q_tensor
 
@@ -184,8 +185,8 @@ class STEUniformActivationQATQuantizer(BasePytorchQATTrainableQuantizer):
         Returns:
             A pytorch inferable quanizer object.
         """
-        _min = self.quantizer_parameters[FQ_MIN].cpu().detach().numpy()
-        _max = self.quantizer_parameters[FQ_MAX].cpu().detach().numpy()
+        _min = self.get_quantizer_variable(FQ_MIN).cpu().detach().numpy()
+        _max = self.get_quantizer_variable(FQ_MAX).cpu().detach().numpy()
 
         return ActivationUniformInferableQuantizer(num_bits=self.num_bits,
                                                    min_range=_min, max_range=_max)

model_compression_toolkit/{qat/common → quantizers_infrastructure}/constants.py

@@ -1,4 +1,4 @@
-# Copyright 2022 Sony Semiconductor Israel, Inc. All rights reserved.
+# Copyright 2023 Sony Semiconductor Israel, Inc. All rights reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -13,6 +13,7 @@
 # limitations under the License.
 # ==============================================================================
 
+# Quantizers constants (for GPTQ, QAT, etc.)
 FQ_MIN = "min"
 FQ_MAX = "max"
 THRESHOLD_TENSOR = "ptq_threshold_tensor"

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/base_inferable_quantizer.py

@@ -15,7 +15,7 @@
 from enum import Enum
 from typing import Any, Dict, List
 
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 
 
 class QuantizationTarget(Enum):

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_all_subclasses.py

@@ -28,5 +28,4 @@ def get_all_subclasses(cls: type) -> Set[type]:
 
     """
 
-    return set(cls.__subclasses__()).union(
-        [s for c in cls.__subclasses__() for s in get_all_subclasses(c)])
+    return set(cls.__subclasses__()).union([s for c in cls.__subclasses__() for s in get_all_subclasses(c)])

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/common/get_quantizers.py

@@ -13,8 +13,8 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common import Logger
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure import QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import QUANTIZATION_TARGET, \
     QUANTIZATION_METHOD
@@ -41,7 +41,7 @@ def get_inferable_quantizer_class(quant_target: QuantizationTarget,
     qat_quantizer_classes = get_all_subclasses(quantizer_base_class)
     filtered_quantizers = list(filter(lambda q_class: getattr(q_class, QUANTIZATION_TARGET) == quant_target and
                                                       getattr(q_class, QUANTIZATION_METHOD) is not None and
-                                                       quant_method in getattr(q_class, QUANTIZATION_METHOD),
+                                                      quant_method in getattr(q_class, QUANTIZATION_METHOD),
                                       qat_quantizer_classes))
 
     if len(filtered_quantizers) != 1:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/load_model.py

@@ -12,8 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.core.common import Logger
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.constants import FOUND_TF
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.get_all_subclasses import get_all_subclasses
 
 if FOUND_TF:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantize_wrapper.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 from typing import Dict, List, Any, Tuple
 from model_compression_toolkit import quantizers_infrastructure as qi
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.logger import Logger
 from model_compression_toolkit.quantizers_infrastructure import BaseInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import WEIGHTS_QUANTIZERS, ACTIVATION_QUANTIZERS, LAYER, STEPS, TRAINING
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/__init__.py

@@ -18,6 +18,10 @@ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructur
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.weights_inferable_quantizers.weights_pot_inferable_quantizer import WeightsPOTInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.weights_inferable_quantizers.weights_symmetric_inferable_quantizer import WeightsSymmetricInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.weights_inferable_quantizers.weights_uniform_inferable_quantizer import WeightsUniformInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.weights_inferable_quantizers.weights_lut_symmetric_inferable_quantizer import WeightsLUTSymmetricInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.weights_inferable_quantizers.weights_lut_pot_inferable_quantizer import WeightsLUTPOTInferableQuantizer
+
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.activation_inferable_quantizers.activation_pot_inferable_quantizer import ActivationPOTInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.activation_inferable_quantizers.activation_symmetric_inferable_quantizer import ActivationSymmetricInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.activation_inferable_quantizers.activation_uniform_inferable_quantizer import ActivationUniformInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.keras.quantizers.activation_inferable_quantizers.activation_lut_pot_inferable_quantizer import ActivationLutPOTInferableQuantizer

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_lut_pot_inferable_quantizer.py

@@ -17,10 +17,10 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.logger import Logger
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.constants import FOUND_TF
 
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure import QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import MULTIPLIER_N_BITS, EPS

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_pot_inferable_quantizer.py

@@ -16,10 +16,10 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.logger import Logger
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.constants import FOUND_TF
 
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure import QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_symmetric_inferable_quantizer.py

@@ -16,9 +16,9 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.constants import FOUND_TF
 
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/activation_inferable_quantizers/activation_uniform_inferable_quantizer.py

@@ -16,9 +16,9 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.logger import Logger
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure import QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.quant_utils import \

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/base_keras_inferable_quantizer.py

@@ -14,7 +14,7 @@
 # ==============================================================================
 from abc import abstractmethod
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.constants import FOUND_TF
 from model_compression_toolkit.quantizers_infrastructure import BaseInferableQuantizer
 
 if FOUND_TF:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/constants.py

@@ -22,3 +22,4 @@ MIN_RANGE = 'min_range'
 MAX_RANGE = 'max_range'
 CHANNEL_AXIS = 'channel_axis'
 INPUT_RANK = 'input_rank'
+CLUSTER_CENTERS = 'cluster_centers'

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_lut_pot_inferable_quantizer.py

@@ -16,8 +16,8 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import MULTIPLIER_N_BITS, EPS

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_lut_symmetric_inferable_quantizer.py

@@ -17,8 +17,8 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import MULTIPLIER_N_BITS, EPS

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_pot_inferable_quantizer.py

@@ -16,8 +16,8 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, QuantizationTarget
 
 if FOUND_TF:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_symmetric_inferable_quantizer.py

@@ -16,9 +16,9 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
+from model_compression_toolkit.constants import FOUND_TF
 
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, QuantizationTarget
 
 if FOUND_TF:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/quantizers/weights_inferable_quantizers/weights_uniform_inferable_quantizer.py

@@ -16,8 +16,8 @@ from typing import List
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TF
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TF
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.quant_utils import \
     adjust_range_to_include_zero

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/keras/validation_functions.py

@@ -16,7 +16,7 @@ from typing import Any
 
 import numpy as np
 
-from model_compression_toolkit.core.common import Logger
+from model_compression_toolkit.logger import Logger
 
 
 def validate_uniform_min_max_ranges(min_range: Any, max_range: Any) -> None:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantize_wrapper.py

@@ -13,8 +13,8 @@
 # limitations under the License.
 # ==============================================================================f
 from typing import List, Union, Any, Dict, Tuple
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.logger import Logger
 from model_compression_toolkit.quantizers_infrastructure import BaseInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants import LAYER, TRAINING
 import inspect
@@ -184,13 +184,11 @@ if FOUND_TORCH:
             return self._weights_vars
 
         def forward(self,
-                    x: torch.Tensor,
                     *args: List[Any],
                     **kwargs: Dict[str, Any]) -> Union[torch.Tensor, List[torch.Tensor]]:
             """
             PytorchQuantizationWrapper forward functions
             Args:
-                x: layer's inputs
                 args: arguments to pass to internal layer.
                 kwargs: key-word dictionary to pass to the internal layer.
 
@@ -218,7 +216,7 @@ if FOUND_TORCH:
             # ----------------------------------
             # Layer operation
             # ----------------------------------
-            outputs = self.layer(x, *args, **kwargs)
+            outputs = self.layer(*args, **kwargs)
 
             # ----------------------------------
             # Quantize all activations
@@ -240,6 +238,18 @@ if FOUND_TORCH:
 
             return outputs
 
+        def get_quantized_weights(self) -> Dict[str, torch.Tensor]:
+            """
+
+            Returns: A dictionary of weights attributes to quantized weights.
+
+            """
+            quantized_weights = {}
+            weights_var = self.get_weights_vars()
+            for name, w, quantizer in weights_var:
+                quantized_weights[name] = quantizer(w)
+            return quantized_weights
+
 else:
     class PytorchQuantizationWrapper(object):
         def __init__(self,

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/__init__.py

@@ -19,6 +19,8 @@ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructur
     import ActivationSymmetricInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.activation_inferable_quantizers.activation_uniform_inferable_quantizer \
     import ActivationUniformInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.activation_inferable_quantizers.activation_lut_pot_inferable_quantizer \
+    import ActivationLutPOTInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.base_pytorch_inferable_quantizer \
     import BasePyTorchInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.weights_inferable_quantizers.weights_pot_inferable_quantizer \
@@ -27,3 +29,7 @@ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructur
     import WeightsSymmetricInferableQuantizer
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.weights_inferable_quantizers.weights_uniform_inferable_quantizer \
     import WeightsUniformInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.weights_inferable_quantizers.weights_lut_symmetric_inferable_quantizer \
+    import WeightsLUTSymmetricInferableQuantizer
+from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.pytorch.quantizers.weights_inferable_quantizers.weights_lut_pot_inferable_quantizer \
+    import WeightsLUTPOTInferableQuantizer

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/{common → pytorch/quantizers/activation_inferable_quantizers}/activation_lut_pot_inferable_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer \
     import mark_quantizer, QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants \

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_pot_inferable_quantizer.py

@@ -15,8 +15,8 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_symmetric_inferable_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/activation_inferable_quantizers/activation_uniform_inferable_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_lut_symmetric_inferable_quantizer.py

@@ -15,8 +15,8 @@
 import numpy as np
 import warnings
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer \
     import mark_quantizer
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_pytorch_inferable_quantizer.py

@@ -14,7 +14,7 @@
 # ==============================================================================
 from abc import abstractmethod
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
+from model_compression_toolkit.constants import FOUND_TORCH
 from model_compression_toolkit.quantizers_infrastructure import BaseInferableQuantizer
 
 if FOUND_TORCH:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_symmetric_inferable_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 
 if FOUND_TORCH:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/base_uniform_inferable_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer
 
 if FOUND_TORCH:

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/constants.py

@@ -21,3 +21,6 @@ PER_CHANNEL = 'per_channel'
 MIN_RANGE = 'min_range'
 MAX_RANGE = 'max_range'
 CHANNEL_AXIS = 'channel_axis'
+CLUSTER_CENTERS = 'cluster_centers'
+MULTIPLIER_N_BITS = 'multiplier_n_bits'
+EPS = 'eps'

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_lut_pot_inferable_quantizer.py

@@ -15,8 +15,8 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer \
     import mark_quantizer, QuantizationTarget
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.constants \

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_lut_symmetric_inferable_quantizer.py

@@ -15,8 +15,8 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer \
     import mark_quantizer, \
     QuantizationTarget

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_pot_inferable_quantizer.py

@@ -15,8 +15,8 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_symmetric_inferable_quantizer.py

@@ -15,8 +15,8 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import mark_quantizer, \
     QuantizationTarget
 

model_compression_toolkit/quantizers_infrastructure/inferable_infrastructure/pytorch/quantizers/weights_inferable_quantizers/weights_uniform_inferable_quantizer.py

@@ -15,9 +15,9 @@
 
 import numpy as np
 
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
-from model_compression_toolkit.core.common.logger import Logger
-from model_compression_toolkit.core.common.target_platform import QuantizationMethod
+from model_compression_toolkit.constants import FOUND_TORCH
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructure.common.base_inferable_quantizer import QuantizationTarget, \
     mark_quantizer