PyPI - mct-nightly - Versions diffs - 2.2.0.20240902.511__tar.gz → 2.2.0.20240904.449__tar.gz - Mend

mct-nightly 2.2.0.20240902.511tar.gz → 2.2.0.20240904.449tar.gz

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Files changed (548) hide show

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: mct-nightly
-Version: 2.2.0.20240902.511
+Version: 2.2.0.20240904.449
 Summary: A Model Compression Toolkit for neural networks
 Home-page: UNKNOWN
 License: UNKNOWN
@@ -57,11 +57,11 @@ Description: # Model Compression Toolkit (MCT)
         Currently, MCT is being tested on various Python, Pytorch and TensorFlow versions:
-        |             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              |
-        |-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-        | Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   |
-        | Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) |
-        | Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) |
+        |             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              | PyTorch 2.4                                                                                                                                                                                                              |
+        |-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+        | Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml)   |
+        | Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml) |
+        | Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml) |

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/README.md RENAMED Viewed

@@ -51,11 +51,11 @@ for hands-on learning. For example:
 Currently, MCT is being tested on various Python, Pytorch and TensorFlow versions:
-|             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              |
-|-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-| Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   |
-| Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) |
-| Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) |
+|             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              | PyTorch 2.4                                                                                                                                                                                                              |
+|-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+| Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml)   |
+| Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml) |
+| Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml) |

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/mct_nightly.egg-info/PKG-INFO RENAMED Viewed

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: mct-nightly
-Version: 2.2.0.20240902.511
+Version: 2.2.0.20240904.449
 Summary: A Model Compression Toolkit for neural networks
 Home-page: UNKNOWN
 License: UNKNOWN
@@ -57,11 +57,11 @@ Description: # Model Compression Toolkit (MCT)
         Currently, MCT is being tested on various Python, Pytorch and TensorFlow versions:
-        |             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              |
-        |-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
-        | Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   |
-        | Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) |
-        | Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) |
+        |             |  PyTorch 2.1                                                                                                                                                                                                               | PyTorch 2.2                                                                                                                                                                                                              | PyTorch 2.3                                                                                                                                                                                                              | PyTorch 2.4                                                                                                                                                                                                              |
+        |-------------|----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
+        | Python 3.9  | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch21.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch22.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch23.yml)   | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python39_pytorch24.yml)   |
+        | Python 3.10 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python310_pytorch24.yml) |
+        | Python 3.11 | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch21.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch22.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch23.yml) | [![Run Tests](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml/badge.svg)](https://github.com/sony/model_optimization/actions/workflows/run_tests_python311_pytorch24.yml) |

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/mct_nightly.egg-info/SOURCES.txt RENAMED Viewed

@@ -421,9 +421,8 @@ model_compression_toolkit/qat/keras/quantizer/ste_rounding/uniform_ste.py
 model_compression_toolkit/qat/pytorch/__init__.py
 model_compression_toolkit/qat/pytorch/quantization_facade.py
 model_compression_toolkit/qat/pytorch/quantizer/__init__.py
-model_compression_toolkit/qat/pytorch/quantizer/base_pytorch_qat_quantizer.py
+model_compression_toolkit/qat/pytorch/quantizer/base_pytorch_qat_weight_quantizer.py
 model_compression_toolkit/qat/pytorch/quantizer/quantization_builder.py
-model_compression_toolkit/qat/pytorch/quantizer/quantizer_utils.py
 model_compression_toolkit/qat/pytorch/quantizer/lsq/__init__.py
 model_compression_toolkit/qat/pytorch/quantizer/lsq/symmetric_lsq.py
 model_compression_toolkit/qat/pytorch/quantizer/lsq/uniform_lsq.py
@@ -502,6 +501,7 @@ model_compression_toolkit/trainable_infrastructure/common/get_quantizer_config.p
 model_compression_toolkit/trainable_infrastructure/common/get_quantizers.py
 model_compression_toolkit/trainable_infrastructure/common/quant_utils.py
 model_compression_toolkit/trainable_infrastructure/common/trainable_quantizer_config.py
+model_compression_toolkit/trainable_infrastructure/common/training_method.py
 model_compression_toolkit/trainable_infrastructure/keras/__init__.py
 model_compression_toolkit/trainable_infrastructure/keras/base_keras_quantizer.py
 model_compression_toolkit/trainable_infrastructure/keras/config_serialization.py
@@ -510,6 +510,15 @@ model_compression_toolkit/trainable_infrastructure/keras/quantize_wrapper.py
 model_compression_toolkit/trainable_infrastructure/keras/quantizer_utils.py
 model_compression_toolkit/trainable_infrastructure/pytorch/__init__.py
 model_compression_toolkit/trainable_infrastructure/pytorch/base_pytorch_quantizer.py
+model_compression_toolkit/trainable_infrastructure/pytorch/quantizer_utils.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/__init__.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/base_activation_quantizer.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/lsq/__init__.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/lsq/symmetric_lsq.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/lsq/uniform_lsq.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/ste/__init__.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/ste/symmetric_ste.py
+model_compression_toolkit/trainable_infrastructure/pytorch/activation_quantizers/ste/uniform_ste.py
 model_compression_toolkit/xquant/__init__.py
 model_compression_toolkit/xquant/common/__init__.py
 model_compression_toolkit/xquant/common/constants.py

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/__init__.py RENAMED Viewed

@@ -27,4 +27,4 @@ from model_compression_toolkit import data_generation
 from model_compression_toolkit import pruning
 from model_compression_toolkit.trainable_infrastructure.keras.load_model import keras_load_quantized_model
-__version__ = "2.2.0.20240902.000511"
+__version__ = "2.2.0.20240904.000449"

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/gptq/pytorch/quantizer/quantization_builder.py RENAMED Viewed

@@ -27,7 +27,6 @@ from mct_quantizers.pytorch.quantizers import BasePyTorchInferableQuantizer
 from model_compression_toolkit.logger import Logger
 from model_compression_toolkit.trainable_infrastructure.common.get_quantizer_config import \
     get_trainable_quantizer_weights_config
-from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_quantizer import BasePytorchQATTrainableQuantizer
 from model_compression_toolkit.trainable_infrastructure.common.get_quantizers import \
     get_trainable_quantizer_class
@@ -35,7 +34,7 @@ from model_compression_toolkit.trainable_infrastructure.common.get_quantizers im
 def quantization_builder(n: common.BaseNode,
                          gptq_config: GradientPTQConfig,
                          kernel_attr: str = None
-                         ) -> Tuple[Dict[str, BasePytorchQATTrainableQuantizer], List[BasePyTorchInferableQuantizer]]:
+                         ) -> Tuple[Dict[str, BasePytorchGPTQTrainableQuantizer], List[BasePyTorchInferableQuantizer]]:
     """
     Build quantizers for a node according to its quantization configuration and
     a global NoOpQuantizeConfig object.

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/__init__.py RENAMED Viewed

@@ -12,7 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from model_compression_toolkit.qat.common.qat_config import QATConfig, TrainingMethod
+from model_compression_toolkit.qat.common.qat_config import QATConfig
 from model_compression_toolkit.qat.keras.quantization_facade import keras_quantization_aware_training_init_experimental, keras_quantization_aware_training_finalize_experimental
-from model_compression_toolkit.qat.pytorch.quantization_facade import pytorch_quantization_aware_training_init_experimental, pytorch_quantization_aware_training_finalize_experimental
+from model_compression_toolkit.qat.pytorch.quantization_facade import pytorch_quantization_aware_training_init_experimental, pytorch_quantization_aware_training_finalize_experimental

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/common/qat_config.py RENAMED Viewed

@@ -14,10 +14,9 @@
 # ==============================================================================
 from typing import Dict
-from enum import Enum
 from model_compression_toolkit.core import common
 from model_compression_toolkit.core.common.framework_info import FrameworkInfo
-from model_compression_toolkit.logger import Logger
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
 def is_qat_applicable(node: common.BaseNode,
@@ -38,23 +37,6 @@ def is_qat_applicable(node: common.BaseNode,
             or node.is_activation_quantization_enabled()
-class TrainingMethod(Enum):
-    """
-    An enum for selecting a QAT training method
-    STE - Standard straight-through estimator. Includes PowerOfTwo, symmetric & uniform quantizers
-    DQA -  DNN Quantization with Attention. Includes a smooth quantization introduces by DQA method
-    LSQ - Learned Step size Quantization. Includes PowerOfTwo, symmetric & uniform quantizers: https://arxiv.org/pdf/1902.08153.pdf
-    """
-    STE = "STE",
-    DQA = "DQA",
-    LSQ = "LSQ"
 class QATConfig:
     """
     QAT configuration class.

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/keras/quantization_facade.py RENAMED Viewed

@@ -24,7 +24,6 @@ from model_compression_toolkit.core.common.mixed_precision.resource_utilization_
 from model_compression_toolkit.core.common.mixed_precision.mixed_precision_quantization_config import \
     MixedPrecisionQuantizationConfig
 from mct_quantizers import KerasActivationQuantizationHolder
-from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper
 from model_compression_toolkit.target_platform_capabilities.target_platform.targetplatform2framework import TargetPlatformCapabilities
 from model_compression_toolkit.core.runner import core_runner
 from model_compression_toolkit.ptq.runner import ptq_runner
@@ -34,6 +33,7 @@ if FOUND_TF:
     from tensorflow.keras.layers import Layer
     from tensorflow.keras.models import Model
+    from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper
     from model_compression_toolkit.core.keras.default_framework_info import DEFAULT_KERAS_INFO
     from model_compression_toolkit.core.keras.keras_implementation import KerasImplementation
     from model_compression_toolkit.core.keras.keras_model_validation import KerasModelValidation

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/keras/quantizer/lsq/symmetric_lsq.py RENAMED Viewed

@@ -20,7 +20,7 @@ import tensorflow as tf
 from tensorflow.python.framework.tensor_shape import TensorShape
 from model_compression_toolkit.constants import SIGNED
-from model_compression_toolkit.qat import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
 from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/keras/quantizer/lsq/uniform_lsq.py RENAMED Viewed

@@ -18,7 +18,7 @@ from tensorflow.python.framework.tensor_shape import TensorShape
 from model_compression_toolkit.constants import RANGE_MIN, RANGE_MAX
 from model_compression_toolkit.trainable_infrastructure.common.constants import FQ_MIN, FQ_MAX
 from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper
-from model_compression_toolkit.qat import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
 from mct_quantizers import mark_quantizer, QuantizationMethod, QuantizationTarget
 from mct_quantizers.keras.quantizers import \

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/keras/quantizer/ste_rounding/symmetric_ste.py RENAMED Viewed

@@ -21,7 +21,7 @@ from tensorflow.python.framework.tensor_shape import TensorShape
 from model_compression_toolkit.constants import SIGNED
 from model_compression_toolkit.trainable_infrastructure.common.constants import FQ_MIN, FQ_MAX
-from model_compression_toolkit.qat import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
 from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/keras/quantizer/ste_rounding/uniform_ste.py RENAMED Viewed

@@ -18,7 +18,7 @@ from tensorflow.python.framework.tensor_shape import TensorShape
 from model_compression_toolkit.constants import RANGE_MIN, RANGE_MAX
 from model_compression_toolkit.trainable_infrastructure.common.constants import FQ_MIN, FQ_MAX
 from model_compression_toolkit.trainable_infrastructure import KerasTrainableQuantizationWrapper
-from model_compression_toolkit.qat import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
 from mct_quantizers import mark_quantizer, QuantizationMethod, QuantizationTarget
 from mct_quantizers.keras.quantizers import \

mct-nightly-2.2.0.20240902.511/model_compression_toolkit/qat/pytorch/quantizer/base_pytorch_qat_quantizer.py → mct-nightly-2.2.0.20240904.449/model_compression_toolkit/qat/pytorch/quantizer/base_pytorch_qat_weight_quantizer.py RENAMED Viewed

@@ -24,23 +24,14 @@ from model_compression_toolkit.trainable_infrastructure.pytorch.base_pytorch_qua
 if FOUND_TORCH:
-    class BasePytorchQATTrainableQuantizer(BasePytorchTrainableQuantizer):
+    class BasePytorchQATWeightTrainableQuantizer(BasePytorchTrainableQuantizer):
         """
-        A base class for trainable Keras quantizer for QAT.
+        A base class for trainable PyTorch weights quantizer for QAT.
         """
-        def __init__(self,
-                     quantization_config: Union[TrainableQuantizerWeightsConfig, TrainableQuantizerActivationConfig]):
-            """
-            Initializes BasePytorchQATTrainableQuantizer object.
-            Args:
-                quantization_config: quantizer config class contains all the information about a quantizer configuration.
-            """
-            super().__init__(quantization_config)
+        pass
 else:  # pragma: no cover
-    class BasePytorchQATTrainableQuantizer(BasePytorchTrainableQuantizer):
+    class BasePytorchQATWeightTrainableQuantizer(BasePytorchTrainableQuantizer):
         def __init__(self,
                      quantization_config: Union[TrainableQuantizerWeightsConfig, TrainableQuantizerActivationConfig]):
             super().__init__(quantization_config)

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/pytorch/quantizer/lsq/symmetric_lsq.py RENAMED Viewed

@@ -18,56 +18,27 @@ import numpy as np
 import torch
 import torch.nn as nn
-from model_compression_toolkit.qat import TrainingMethod
 from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
 from mct_quantizers import PytorchQuantizationWrapper
 from model_compression_toolkit.qat.common import THRESHOLD_TENSOR
 from model_compression_toolkit import constants as C
-from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_quantizer import BasePytorchQATTrainableQuantizer
+from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_weight_quantizer import BasePytorchQATWeightTrainableQuantizer
 from mct_quantizers.common.base_inferable_quantizer import mark_quantizer, QuantizationTarget
 from model_compression_toolkit.core.pytorch.utils import to_torch_tensor
-from model_compression_toolkit.qat.pytorch.quantizer.quantizer_utils import ste_round, grad_scale
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure.pytorch.quantizer_utils import symmetric_lsq_quantizer
 from mct_quantizers.pytorch.quantizers import \
-    WeightsPOTInferableQuantizer, WeightsSymmetricInferableQuantizer, ActivationPOTInferableQuantizer, \
-    ActivationSymmetricInferableQuantizer
+    WeightsPOTInferableQuantizer, WeightsSymmetricInferableQuantizer
 from model_compression_toolkit.trainable_infrastructure.common.trainable_quantizer_config import \
-    TrainableQuantizerWeightsConfig, TrainableQuantizerActivationConfig
+    TrainableQuantizerWeightsConfig
 from model_compression_toolkit.trainable_infrastructure.common.base_trainable_quantizer import VariableGroup
-def symmetric_lsq_quantizer(x: nn.Parameter,
-                          thresholds: nn.Parameter,
-                          num_bits: int,
-                          sign: bool,
-                          min_int: int,
-                          max_int: int,
-                          scale_factor: float) -> Union[nn.Parameter, torch.Tensor]:
-    """
-    Symmetric quantizer according to LSQ algorithm: https://arxiv.org/pdf/1902.08153.pdf
-    Args:
-        x: input to quantize
-        thresholds: thresholds of quantization levels
-        num_bits: number of bits for quantization
-        sign: whether x is signed or not
-        min_int: min clipping integer value
-        max_int: max clipping integer value
-        scale_factor: grad scale of LSQ algorithm
-    Returns:
-        A quantized tensor
-    """
-    delta = thresholds / (2 ** (num_bits - int(sign)))
-    delta_scaled = grad_scale(delta, scale_factor)
-    rounded = ste_round(x / delta_scaled)
-    clipped = torch.clip(rounded, min=min_int, max=max_int)
-    quantized = delta_scaled * clipped
-    return quantized
 @mark_quantizer(quantization_target=QuantizationTarget.Weights,
                 quantization_method=[QuantizationMethod.POWER_OF_TWO, QuantizationMethod.SYMMETRIC],
                 identifier=TrainingMethod.LSQ)
-class LSQWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
+class LSQWeightQATQuantizer(BasePytorchQATWeightTrainableQuantizer):
     """
     Trainable constrained quantizer to quantize layer's weights.
     """
@@ -145,84 +116,3 @@ class LSQWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
                                                       threshold=threshold_values.tolist(),
                                                       per_channel=self.quantization_config.weights_per_channel_threshold,
                                                       channel_axis=self.quantization_config.weights_channels_axis)
-@mark_quantizer(quantization_target=QuantizationTarget.Activation,
-                quantization_method=[QuantizationMethod.POWER_OF_TWO, QuantizationMethod.SYMMETRIC],
-                identifier=TrainingMethod.LSQ)
-class LSQActivationQATQuantizer(BasePytorchQATTrainableQuantizer):
-    """
-    Trainable constrained quantizer to quantize layer activations.
-    """
-    def __init__(self, quantization_config: TrainableQuantizerActivationConfig):
-        """
-        Initialize a LSQActivationQATQuantizer object with parameters to use
-        for symmetric or power of two quantization.
-        Args:
-            quantization_config: trainable quantizer config class
-        """
-        super().__init__(quantization_config)
-        self.power_of_two = quantization_config.activation_quantization_method == QuantizationMethod.POWER_OF_TWO
-        self.sign = quantization_config.activation_quantization_params['is_signed']
-        self.threshold_values = np.array([quantization_config.activation_quantization_params[C.THRESHOLD]])
-        self.num_bits = quantization_config.activation_n_bits
-        n_pos_bits = self.num_bits - int(self.sign)
-        self.min_int = -int(self.sign) * (2 ** n_pos_bits)
-        self.max_int = (2 ** n_pos_bits) - 1
-    def initialize_quantization(self,
-                                tensor_shape: torch.Size,
-                                name: str,
-                                layer: PytorchQuantizationWrapper):
-        """
-        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, nn.Parameter(to_torch_tensor(self.threshold_values), requires_grad=True))
-        # save the quantizer added parameters for later calculations
-        self.add_quantizer_variable(THRESHOLD_TENSOR, layer.get_parameter(name), VariableGroup.QPARAMS)
-    def __call__(self,
-                 inputs: torch.Tensor,
-                 training: bool = True) -> torch.Tensor:
-        """
-        Quantize a tensor.
-        Args:
-            inputs: Input tensor to quantize.
-            training: Whether the graph is in training mode.
-        Returns:
-            The quantized tensor.
-        """
-        thresholds = self.get_quantizer_variable(THRESHOLD_TENSOR)
-        n_channels = inputs.shape[1]
-        scale_factor = 1.0 / np.sqrt(self.max_int * n_channels)
-        inputs_quantized = symmetric_lsq_quantizer(inputs, thresholds, self.num_bits, self.sign, self.min_int, self.max_int, scale_factor)
-        return inputs_quantized
-    def convert2inferable(self) -> Union[ActivationPOTInferableQuantizer, ActivationSymmetricInferableQuantizer]:
-        """
-        Convert quantizer to inferable quantizer.
-        Returns:
-            A pytorch inferable quanizer object.
-        """
-        threshold_values = self.get_quantizer_variable(THRESHOLD_TENSOR).cpu().detach().numpy()
-        if self.power_of_two:
-            pot_threshold = np.power(2.0, np.ceil(np.log2(threshold_values)))
-            return ActivationPOTInferableQuantizer(num_bits=self.num_bits,
-                                                   threshold=pot_threshold.tolist(),
-                                                   signed=self.sign)
-        else:
-            return ActivationSymmetricInferableQuantizer(num_bits=self.num_bits,
-                                                         threshold=threshold_values.tolist(),
-                                                         signed=self.sign)

{mct-nightly-2.2.0.20240902.511 → mct-nightly-2.2.0.20240904.449}/model_compression_toolkit/qat/pytorch/quantizer/lsq/uniform_lsq.py RENAMED Viewed

@@ -12,66 +12,32 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ==============================================================================
-from typing import Union
 import numpy as np
 import torch
 import torch.nn as nn
-from model_compression_toolkit.constants import RANGE_MAX, RANGE_MIN
-from model_compression_toolkit.trainable_infrastructure.common.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 mct_quantizers import QuantizationTarget, PytorchQuantizationWrapper
-from model_compression_toolkit import constants as C
-from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_quantizer import BasePytorchQATTrainableQuantizer
 from mct_quantizers import mark_quantizer
-from model_compression_toolkit.qat.pytorch.quantizer.quantizer_utils import ste_round, grad_scale
-from model_compression_toolkit.core.pytorch.utils import to_torch_tensor
 from mct_quantizers.pytorch.quantizers import \
-    WeightsUniformInferableQuantizer, ActivationUniformInferableQuantizer
-from model_compression_toolkit.trainable_infrastructure.common.trainable_quantizer_config import \
-    TrainableQuantizerWeightsConfig, TrainableQuantizerActivationConfig
+    WeightsUniformInferableQuantizer
+from model_compression_toolkit.constants import RANGE_MAX, RANGE_MIN
+from model_compression_toolkit.trainable_infrastructure.common.constants import FQ_MIN, FQ_MAX
+from model_compression_toolkit.trainable_infrastructure import TrainingMethod
+from model_compression_toolkit.trainable_infrastructure.pytorch.quantizer_utils import uniform_lsq_quantizer
 from model_compression_toolkit.trainable_infrastructure.common.base_trainable_quantizer import VariableGroup
-from model_compression_toolkit.qat.pytorch.quantizer.quantizer_utils import adjust_range_to_include_zero
+from model_compression_toolkit.trainable_infrastructure.common.trainable_quantizer_config import \
+    TrainableQuantizerWeightsConfig
+from model_compression_toolkit.target_platform_capabilities.target_platform import QuantizationMethod
+from model_compression_toolkit.core.pytorch.utils import to_torch_tensor
 from model_compression_toolkit.core.common.quantization.quantizers.quantizers_helpers import fix_range_to_include_zero
-def uniform_lsq_quantizer(x: nn.Parameter,
-                      min_range: nn.Parameter,
-                      max_range: nn.Parameter,
-                      num_bits: int,
-                      min_int: int,
-                      max_int: int,
-                      scale_factor: float) -> Union[nn.Parameter, torch.Tensor]:
-    """
-    Uniform quantizer according to LSQ algorithm: https://arxiv.org/pdf/1902.08153.pdf
-    Args:
-        x: input to quantize
-        min_range: min range of quantization values
-        max_range: min range of quantization values
-        num_bits: number of bits for quantization
-        min_int: min clipping integer value
-        max_int: max clipping integer value
-        scale_factor: grad scale of LSQ algorithm
-    Returns:
-        A quantized tensor
-    """
-    a, b = adjust_range_to_include_zero(min_range, max_range, num_bits)
-    delta = (b - a) / (2 ** num_bits - 1)
-    delta_scaled = grad_scale(delta, scale_factor)
-    rounded = ste_round((x - a) / delta_scaled)
-    clipped = torch.clip(rounded, min=min_int, max=max_int)
-    quantized = delta_scaled * clipped + a
-    return quantized
+from model_compression_toolkit.qat.pytorch.quantizer.base_pytorch_qat_weight_quantizer import BasePytorchQATWeightTrainableQuantizer
 @mark_quantizer(quantization_target=QuantizationTarget.Weights,
                 quantization_method=[QuantizationMethod.UNIFORM],
                 identifier=TrainingMethod.LSQ)
-class LSQUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
+class LSQUniformWeightQATQuantizer(BasePytorchQATWeightTrainableQuantizer):
     """
     Trainable constrained quantizer to quantize layer's weights.
     """
@@ -145,79 +111,3 @@ class LSQUniformWeightQATQuantizer(BasePytorchQATTrainableQuantizer):
                                                 max_range=max_range.tolist(),
                                                 per_channel=self.quantization_config.weights_per_channel_threshold,
                                                 channel_axis=self.quantization_config.weights_channels_axis)
-@mark_quantizer(quantization_target=QuantizationTarget.Activation,
-                quantization_method=[QuantizationMethod.UNIFORM],
-                identifier=TrainingMethod.LSQ)
-class LSQUniformActivationQATQuantizer(BasePytorchQATTrainableQuantizer):
-    """
-    Trainable constrained quantizer to quantize layer activations.
-    """
-    def __init__(self, quantization_config: TrainableQuantizerActivationConfig):
-        """
-        Initialize a LSQUniformActivationQATQuantizer object with parameters to use
-        for uniform quantization.
-        Args:
-            quantization_config: trainable quantizer config class
-        """
-        super().__init__(quantization_config)
-        self.num_bits = self.quantization_config.activation_n_bits
-        self.min_int = 0
-        self.max_int = 2 ** self.num_bits - 1
-        self.min_range = np.array([quantization_config.activation_quantization_params[C.RANGE_MIN]])
-        self.max_range = np.array([quantization_config.activation_quantization_params[C.RANGE_MAX]])
-    def initialize_quantization(self,
-                                tensor_shape: torch.Size,
-                                name: str,
-                                layer: PytorchQuantizationWrapper):
-        """
-        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), requires_grad=True))
-        layer.register_parameter(name+"_"+FQ_MAX, nn.Parameter(to_torch_tensor(self.max_range), requires_grad=True))
-        # Save the quantizer parameters for later calculations
-        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,
-                 training: bool = True) -> torch.Tensor:
-        """
-        Quantize a tensor.
-        Args:
-            inputs: Input tensor to quantize.
-            training: Whether the graph is in training mode.
-        Returns:
-            The quantized tensor.
-        """
-        min_range = self.get_quantizer_variable(FQ_MIN)
-        max_range = self.get_quantizer_variable(FQ_MAX)
-        n_channels = inputs.shape[1]
-        scale_factor = 1.0 / np.sqrt(self.max_int * n_channels)
-        inputs_quantized = uniform_lsq_quantizer(inputs, min_range, max_range, self.num_bits, self.min_int, self.max_int, scale_factor)
-        return inputs_quantized
-    def convert2inferable(self) -> ActivationUniformInferableQuantizer:
-        """
-        Convert quantizer to inferable quantizer.
-        Returns:
-            A pytorch inferable quanizer object.
-        """
-        min_range = self.get_quantizer_variable(FQ_MIN).cpu().detach().numpy()
-        max_range = self.get_quantizer_variable(FQ_MAX).cpu().detach().numpy()
-        min_range, max_range = fix_range_to_include_zero(min_range, max_range, self.num_bits)
-        return ActivationUniformInferableQuantizer(num_bits=self.num_bits,
-                                                   min_range=min_range.tolist(),
-                                                   max_range=max_range.tolist())

mct-nightly 2.2.0.20240902.511__tar.gz → 2.2.0.20240904.449__tar.gz

mct-nightly 2.2.0.20240902.511tar.gz → 2.2.0.20240904.449tar.gz