mct-nightly 1.11.0.20240110.post414

A Model Compression Toolkit for neural networks

Model Compression Toolkit (MCT)

Model Compression Toolkit (MCT) is an open-source project for neural network model optimization under efficient, constrained hardware.

This project provides researchers, developers, and engineers tools for optimizing and deploying state-of-the-art neural networks on efficient hardware.

Specifically, this project aims to apply quantization to compress neural networks.

MCT is developed by researchers and engineers working at Sony Semiconductor Israel.

Table of Contents

• Getting Started
• Supported features
• Results
• Contributions
• License

Getting Started

This section provides an installation and a quick starting guide.

Installation

To install the latest stable release of MCT, run the following command:

pip install model-compression-toolkit

For installing the nightly version or installing from source, refer to the installation guide.

Quick start & tutorials

For an example of how to use MCT with TensorFlow or PyTorch on various models and tasks, check out the quick-start page and the results CSV.

In addition, a set of notebooks are provided for an easy start. For example:

• MobileNet with Tensorflow.
• MobileNetV2 with PyTorch.
• Data Generation for ResNet18 with PyTorch.

Supported Versions

Currently, MCT is being tested on various Python, Pytorch and TensorFlow versions:

	PyTorch 1.13	PyTorch 2.0	PyTorch 2.1
Python 3.9
Python 3.10
Python 3.11

	TensorFlow 2.12	TensorFlow 2.13	TensorFlow 2.14
Python 3.9
Python 3.10
Python 3.11

Supported Features

MCT offers a range of powerful features to optimize neural network models for efficient deployment. These supported features include:

Data Generation

MCT provides tools for generating synthetic images based on the statistics stored in a model's batch normalization layers. These generated images are valuable for various compression tasks where image data is required, such as quantization and pruning. You can customize data generation configurations to suit your specific needs. Go to the Data Generation page.

Quantization

MCT supports different quantization methods:

• Post-training quantization (PTQ): Keras API, PyTorch API
• Gradient-based post-training quantization (GPTQ): Keras API, PyTorch API
• Quantization-aware training (QAT)*

Quantization Method	Complexity	Computational Cost
PTQ	Low	Low (order of minutes)
GPTQ (parameters fine-tuning using gradients)	Mild	Mild (order of 2-3 hours)
QAT	High	High (order of 12-36 hours)

In addition, MCT supports different quantization schemes for quantizing weights and activations:

• Power-Of-Two (hardware-friendly quantization [1])
• Symmetric
• Uniform

Main features:

• Graph optimizations: Transforming the model to an equivalent (yet, more efficient) model (for example, batch-normalization layer folding to its preceding linear layer).
• Quantization parameter search: Different methods can be used to minimize the expected added quantization-noise during thresholds search (by default, we use Mean-Square-Error, but other metrics can be used such as No-Clipping, Mean-Average-Error, and more).
• Advanced quantization algorithms: To prevent a performance degradation some algorithms are applied such as:
  • Shift negative correction: Symmetric activation quantization can hurt the model's performance when some layers output both negative and positive activations, but their range is asymmetric. For more details please visit [1].
  • Outliers filtering: Computing z-score for activation statistics to detect and remove outliers.
• Clustering: Using non-uniform quantization grid to quantize the weights and activations to match their distributions.*
• Mixed-precision search: Assigning quantization bit-width per layer (for weights/activations), based on the layer's sensitivity to different bit-widths.
• Visualization: You can use TensorBoard to observe useful information for troubleshooting the quantized model's performance (for example, the model in different phases of the quantization, collected statistics, similarity between layers of the float and quantized model and bit-width configuration for mixed-precision quantization). For more details, please read the visualization documentation.
• Target Platform Capabilities: The Target Platform Capabilities (TPC) describes the target platform (an edge device with dedicated hardware). For more details, please read the TPC README.

Enhanced Post-Training Quantization (EPTQ)

As part of the GPTQ we provide an advanced optimization algorithm called EPTQ.

The specifications of the algorithm are detailed in the paper: "EPTQ: Enhanced Post-Training Quantization via Label-Free Hessian" [4].

More details on the how to use EPTQ via MCT can be found in the EPTQ guidelines.

Experimental features

Some features are experimental and subject to future changes.

For more details, we highly recommend visiting our project website where experimental features are mentioned as experimental.

Results

Keras

Graph of MobileNetV2 accuracy on ImageNet vs average bit-width of weights, using single-precision quantization, mixed-precision quantization, and mixed-precision quantization with GPTQ.

For more results, please see [1]

Pytorch

We quantized classification networks from the torchvision library. In the following table we present the ImageNet validation results for these models:

Network Name	Float Accuracy	8Bit Accuracy	Data-Free 8Bit Accuracy
MobileNet V2 [3]	71.886	71.444	71.29
ResNet-18 [3]	69.86	69.63	69.53
SqueezeNet 1.1 [3]	58.128	57.678

For more results, please refer to quick start.

Structured Pruning

MCT introduces a structured and hardware-aware model pruning. This pruning technique is designed to compress models for specific hardware architectures, taking into account the target platform's Single Instruction, Multiple Data (SIMD) capabilities. By pruning groups of channels (SIMD groups), our approach not only reduces model size and complexity, but ensures that better utilization of channels is in line with the SIMD architecture for a target KPI of weights memory footprint.

Note: Currently, only Keras models pruning is supported.

Results

Results for applying pruning to reduce the parameters of the following models by 50%:

Model	Dense Model Accuracy	Pruned Model Accuracy
ResNet50 [2]	75.1	72.4
DenseNet121 [2]	75.0	71.15

Contributions

MCT aims at keeping a more up-to-date fork and welcomes contributions from anyone.

*You will find more information about contributions in the Contribution guide.

License

Apache License 2.0.

References

[1] Habi, H.V., Peretz, R., Cohen, E., Dikstein, L., Dror, O., Diamant, I., Jennings, R.H. and Netzer, A., 2021. HPTQ: Hardware-Friendly Post Training Quantization. arXiv preprint.

[2] Keras Applications

[3] TORCHVISION.MODELS

[4] Gordon, O., Habi, H. V., & Netzer, A., 2023. EPTQ: Enhanced Post-Training Quantization via Label-Free Hessian. arXiv preprint