mct-nightly 1.8.0.22042023.post414__py3-none-any.whl → 1.8.0.22052023.post408__py3-none-any.whl

model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/common/get_quantizers.py

@@ -12,11 +12,9 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from typing import Union
+from typing import Union, Any
 
-from model_compression_toolkit.gptq import RoundingType
-from model_compression_toolkit import TrainingMethod
-from model_compression_toolkit.core.common import Logger
+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 \
@@ -26,7 +24,7 @@ from model_compression_toolkit.quantizers_infrastructure.inferable_infrastructur
 
 
 def get_trainable_quantizer_class(quant_target: QuantizationTarget,
-                                  quantizer_type: Union[TrainingMethod, RoundingType],
+                                  quantizer_type: Union[Any, Any],
                                   quant_method: QuantizationMethod,
                                   quantizer_base_class: type) -> type:
     """

model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/keras/base_keras_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 from typing import Dict, Any, Union, List
 
-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.trainable_infrastructure.common.base_trainable_quantizer import VariableGroup
 from model_compression_toolkit.quantizers_infrastructure.trainable_infrastructure.common.base_trainable_quantizer import BaseTrainableQuantizer, VAR, GROUP
 from model_compression_toolkit.quantizers_infrastructure import TrainableQuantizerWeightsConfig, \

model_compression_toolkit/quantizers_infrastructure/trainable_infrastructure/pytorch/base_pytorch_quantizer.py

@@ -14,8 +14,8 @@
 # ==============================================================================
 from typing import Union, List
 
-from model_compression_toolkit.core.common.logger import Logger
-from model_compression_toolkit.core.common.constants import FOUND_TORCH
+from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.constants import FOUND_TORCH
 from model_compression_toolkit.quantizers_infrastructure.trainable_infrastructure.common.base_trainable_quantizer import VariableGroup
 from model_compression_toolkit.quantizers_infrastructure.trainable_infrastructure.common.base_trainable_quantizer import BaseTrainableQuantizer, VAR, GROUP
 from model_compression_toolkit.quantizers_infrastructure import TrainableQuantizerWeightsConfig, \

model_compression_toolkit/target_platform_capabilities/constants.py

@@ -0,0 +1,27 @@
+# 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.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+
+# TP Model constants
+OPS_SET_LIST = 'ops_set_list'
+
+# Version
+LATEST = 'latest'
+
+
+# Supported TP models names:
+DEFAULT_TP_MODEL = 'default'
+IMX500_TP_MODEL = 'imx500'
+TFLITE_TP_MODEL = 'tflite'
+QNNPACK_TP_MODEL = 'qnnpack'

model_compression_toolkit/target_platform_capabilities/target_platform/current_tp_model.py

@@ -13,7 +13,7 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.logger import Logger
 
 def get_current_tp_model():
     """

model_compression_toolkit/target_platform_capabilities/target_platform/operators.py

@@ -14,7 +14,7 @@
 # ==============================================================================
 from typing import Dict, Any
 
-from model_compression_toolkit.core.common.constants import OPS_SET_LIST
+from model_compression_toolkit.target_platform_capabilities.constants import OPS_SET_LIST
 from model_compression_toolkit.target_platform_capabilities.target_platform.target_platform_model_component import TargetPlatformModelComponent
 from model_compression_toolkit.target_platform_capabilities.target_platform.current_tp_model import _current_tp_model
 from model_compression_toolkit.target_platform_capabilities.target_platform.op_quantization_config import QuantizationConfigOptions

model_compression_toolkit/target_platform_capabilities/target_platform/quantization_format.py

@@ -0,0 +1,20 @@
+# 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.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+from enum import Enum
+
+
+class QuantizationFormat(Enum):
+    FAKELY_QUANT = 0
+    INT8 = 1

model_compression_toolkit/target_platform_capabilities/target_platform/target_platform_model.py

@@ -24,8 +24,8 @@ from model_compression_toolkit.target_platform_capabilities.target_platform.targ
 from model_compression_toolkit.target_platform_capabilities.target_platform.op_quantization_config import OpQuantizationConfig, \
     QuantizationConfigOptions
 from model_compression_toolkit.target_platform_capabilities.target_platform.operators import OperatorsSetBase
-from model_compression_toolkit.core.common.immutable import ImmutableClass
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.target_platform_capabilities.immutable import ImmutableClass
+from model_compression_toolkit.logger import Logger
 
 
 def get_default_quantization_config_options() -> QuantizationConfigOptions:
@@ -223,3 +223,12 @@ class TargetPlatformModel(ImmutableClass):
 
         """
         pprint.pprint(self.get_info(), sort_dicts=False)
+
+    def set_quantization_format(self,
+                                quantization_format: Any):
+        """
+        Set quantization format.
+        Args:
+            quantization_format: A quantization format (fake-quant, int8 etc.) from enum QuantizationFormat.
+        """
+        self.quantization_format = quantization_format

model_compression_toolkit/target_platform_capabilities/target_platform/targetplatform2framework/attribute_filter.py

@@ -16,7 +16,7 @@
 import operator
 from typing import Any, Callable, Dict
 
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.logger import Logger
 
 
 class Filter:

model_compression_toolkit/target_platform_capabilities/target_platform/targetplatform2framework/layer_filter_params.py

@@ -13,9 +13,7 @@
 # limitations under the License.
 # ==============================================================================
 
-from typing import Any, Dict
-
-from model_compression_toolkit.core.common.graph.base_node import BaseNode
+from typing import Any
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.attribute_filter import AttributeFilter
 
 
@@ -87,34 +85,34 @@ class LayerFilterParams:
         params.extend([str(c) for c in self.conditions])
         params_str = ', '.join(params)
         return f'{self.layer.__name__}({params_str})'
-
-    def match(self,
-              node: BaseNode) -> bool:
-        """
-        Check if a node matches the layer, conditions and keyword-arguments of
-        the LayerFilterParams.
-
-        Args:
-            node: Node to check if matches to the LayerFilterParams properties.
-
-        Returns:
-            Whether the node matches to the LayerFilterParams properties.
-        """
-        # Check the node has the same type as the layer in LayerFilterParams
-        if self.layer != node.type:
-            return False
-
-        # Get attributes from node to filter
-        layer_config = node.framework_attr
-        if hasattr(node, "op_call_kwargs"):
-            layer_config.update(node.op_call_kwargs)
-
-        for attr, value in self.kwargs.items():
-            if layer_config.get(attr) != value:
-                return False
-
-        for c in self.conditions:
-            if not c.match(layer_config):
-                return False
-
-        return True
+    #
+    # def match(self,
+    #           node: BaseNode) -> bool:
+    #     """
+    #     Check if a node matches the layer, conditions and keyword-arguments of
+    #     the LayerFilterParams.
+    #
+    #     Args:
+    #         node: Node to check if matches to the LayerFilterParams properties.
+    #
+    #     Returns:
+    #         Whether the node matches to the LayerFilterParams properties.
+    #     """
+    #     # Check the node has the same type as the layer in LayerFilterParams
+    #     if self.layer != node.type:
+    #         return False
+    #
+    #     # Get attributes from node to filter
+    #     layer_config = node.framework_attr
+    #     if hasattr(node, "op_call_kwargs"):
+    #         layer_config.update(node.op_call_kwargs)
+    #
+    #     for attr, value in self.kwargs.items():
+    #         if layer_config.get(attr) != value:
+    #             return False
+    #
+    #     for c in self.conditions:
+    #         if not c.match(layer_config):
+    #             return False
+    #
+    #     return True

model_compression_toolkit/target_platform_capabilities/target_platform/targetplatform2framework/operations_to_layers.py

@@ -15,10 +15,10 @@
 
 from typing import List, Any
 
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.logger import Logger
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.current_tpc import  _current_tpc
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.target_platform_capabilities_component import TargetPlatformCapabilitiesComponent
-from model_compression_toolkit.target_platform_capabilities.target_platform.operators import OperatorsSet, OperatorSetConcat, \
+from model_compression_toolkit.target_platform_capabilities.target_platform.operators import OperatorSetConcat, \
     OperatorsSetBase
 
 

model_compression_toolkit/target_platform_capabilities/target_platform/targetplatform2framework/target_platform_capabilities.py

@@ -18,16 +18,15 @@ import itertools
 import pprint
 from typing import List, Any, Dict, Tuple
 
-from model_compression_toolkit.core.common.logger import Logger
+from model_compression_toolkit.logger import Logger
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.operations_to_layers import \
     OperationsToLayers, OperationsSetToLayers
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.target_platform_capabilities_component import TargetPlatformCapabilitiesComponent
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.layer_filter_params import LayerFilterParams
-from model_compression_toolkit.core.common.immutable import ImmutableClass
-from model_compression_toolkit.core.common.graph.base_node import BaseNode
+from model_compression_toolkit.target_platform_capabilities.immutable import ImmutableClass
 from model_compression_toolkit.target_platform_capabilities.target_platform.op_quantization_config import QuantizationConfigOptions, \
     OpQuantizationConfig
-from model_compression_toolkit.target_platform_capabilities.target_platform.operators import OperatorsSet, OperatorsSetBase
+from model_compression_toolkit.target_platform_capabilities.target_platform.operators import OperatorsSetBase
 from model_compression_toolkit.target_platform_capabilities.target_platform.target_platform_model import TargetPlatformModel
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework.current_tpc import _current_tpc
 
@@ -163,26 +162,6 @@ class TargetPlatformCapabilities(ImmutableClass):
         """
         return self.tp_model.get_default_op_quantization_config()
 
-    def get_qco_by_node(self,
-                        node: BaseNode) -> QuantizationConfigOptions:
-        """
-        Get the QuantizationConfigOptions of a node in a graph according
-        to the mappings from layers/LayerFilterParams to the OperatorsSet in the TargetPlatformModel.
-
-        Args:
-            node: Node from graph to get its QuantizationConfigOptions.
-
-        Returns:
-            QuantizationConfigOptions of the node.
-        """
-        if node is None:
-            Logger.error(f'Can not retrieve QC options for None node')  # pragma: no cover
-        for fl, qco in self.filterlayer2qco.items():
-            if fl.match(node):
-                return qco
-        if node.type in self.layer2qco:
-            return self.layer2qco.get(node.type)
-        return self.tp_model.default_qco
 
     def _get_config_options_mapping(self) -> Tuple[Dict[Any, QuantizationConfigOptions],
                                                    Dict[LayerFilterParams, QuantizationConfigOptions]]:

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/latest/__init__.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH
 
 from model_compression_toolkit.target_platform_capabilities.tpc_models.default_tpc.v4.tp_model import get_tp_model, generate_tp_model, get_op_quantization_configs
 

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/target_platform_capabilities.py

@@ -13,7 +13,9 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH, LATEST
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH
+from model_compression_toolkit.target_platform_capabilities.constants import LATEST
+
 
 ###############################
 # Build Tensorflow TPC models

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v1/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -106,6 +109,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                          tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v2/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -106,6 +109,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                          tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v3/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -113,6 +116,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                          tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v3_lut/tp_model.py

@@ -15,8 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel, \
-    QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel, QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -115,6 +117,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                         tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v4/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -113,6 +116,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                          tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v4_lut/tp_model.py

@@ -15,8 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel, \
-    QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel, QuantizationMethod
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -115,6 +117,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                         tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/default_tpc/v5/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -113,6 +116,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                          tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/get_target_platform_capabilities.py

@@ -23,9 +23,7 @@ from model_compression_toolkit.target_platform_capabilities.tpc_models.tflite_tp
     tpc_dict as tflite_tpc_dict
 from model_compression_toolkit.target_platform_capabilities.tpc_models.qnnpack_tpc.target_platform_capabilities import \
     tpc_dict as qnnpack_tpc_dict
-from model_compression_toolkit.core.keras.constants import DEFAULT_TP_MODEL, IMX500_TP_MODEL, TFLITE_TP_MODEL, \
-    QNNPACK_TP_MODEL
-from model_compression_toolkit.core.common.constants import LATEST
+from model_compression_toolkit.target_platform_capabilities.constants import DEFAULT_TP_MODEL, IMX500_TP_MODEL, TFLITE_TP_MODEL, QNNPACK_TP_MODEL,  LATEST
 
 tpc_dict = {DEFAULT_TP_MODEL: default_tpc_dict,
             IMX500_TP_MODEL: imx500_tpc_dict,

model_compression_toolkit/target_platform_capabilities/tpc_models/imx500_tpc/latest/__init__.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH
 from model_compression_toolkit.target_platform_capabilities.tpc_models.imx500_tpc.v1.tp_model import get_tp_model, generate_tp_model, \
     get_op_quantization_configs
 if FOUND_TF:

model_compression_toolkit/target_platform_capabilities/tpc_models/imx500_tpc/target_platform_capabilities.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH, LATEST
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH
+from model_compression_toolkit.target_platform_capabilities.constants import LATEST
 
 ###############################
 # Build Tensorflow TPC models

model_compression_toolkit/target_platform_capabilities/tpc_models/imx500_tpc/v1/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -113,6 +116,9 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         # be used for operations that will be attached to this set's label.
         # Otherwise, it will be a configure-less set (used in fusing):
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
         # May suit for operations like: Dropout, Reshape, etc.
         tp.OperatorsSet("NoQuantization",
                         tp.get_default_quantization_config_options().clone_and_edit(

model_compression_toolkit/target_platform_capabilities/tpc_models/qnnpack_tpc/latest/__init__.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH
 from model_compression_toolkit.target_platform_capabilities.tpc_models.qnnpack_tpc.v1.tp_model import get_tp_model, generate_tp_model, get_op_quantization_configs
 if FOUND_TF:
     from model_compression_toolkit.target_platform_capabilities.tpc_models.qnnpack_tpc.v1.tpc_keras import get_keras_tpc as get_keras_tpc_latest

model_compression_toolkit/target_platform_capabilities/tpc_models/qnnpack_tpc/target_platform_capabilities.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH, LATEST
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH
+from model_compression_toolkit.target_platform_capabilities.constants import LATEST
 
 
 ###############################

model_compression_toolkit/target_platform_capabilities/tpc_models/qnnpack_tpc/v1/tp_model.py

@@ -15,7 +15,10 @@
 from typing import List, Tuple
 
 import model_compression_toolkit as mct
-from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform import OpQuantizationConfig, \
+    TargetPlatformModel
+from model_compression_toolkit.target_platform_capabilities.target_platform.quantization_format import \
+    QuantizationFormat
 
 tp = mct.target_platform
 
@@ -120,4 +123,7 @@ def generate_tp_model(default_config: OpQuantizationConfig,
         tp.Fusing([conv, relu])
         tp.Fusing([linear, relu])
 
+        # Set quantization format to fakely quant
+        generated_tpc.set_quantization_format(QuantizationFormat.FAKELY_QUANT)
+
     return generated_tpc

model_compression_toolkit/target_platform_capabilities/tpc_models/tflite_tpc/latest/__init__.py

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH
 from model_compression_toolkit.target_platform_capabilities.tpc_models.tflite_tpc.v1.tp_model import get_tp_model, generate_tp_model, get_op_quantization_configs
 if FOUND_TF:
     from model_compression_toolkit.target_platform_capabilities.tpc_models.tflite_tpc.v1.tpc_keras import get_keras_tpc as get_keras_tpc_latest

model_compression_toolkit/target_platform_capabilities/tpc_models/tflite_tpc/target_platform_capabilities.py

@@ -13,7 +13,8 @@
 # limitations under the License.
 # ==============================================================================
 
-from model_compression_toolkit.core.common.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH, LATEST
+from model_compression_toolkit.constants import FOUND_TF, FOUND_TORCH, TENSORFLOW, PYTORCH
+from model_compression_toolkit.target_platform_capabilities.constants import LATEST
 
 
 ###############################