Introduction to Intel® Neural Compressor

Intel® Neural Compressor (formerly known as Intel® Low Precision Optimization Tool) is an open-source Python library running on Intel CPUs and GPUs, which delivers unified interfaces across multiple deep learning frameworks for popular network compression technologies, such as quantization, pruning, knowledge distillation. This tool supports automatic accuracy-driven tuning strategies to help user quickly find out the best quantized model. It also implements different weight pruning algorithms to generate pruned model with predefined sparsity goal and supports knowledge distillation to distill the knowledge from the teacher model to the student model.

Note

GPU support is under development.

Visit the Intel® Neural Compressor online document website at: https://intel.github.io/neural-compressor.

Infrastructure

Intel® Neural Compressor features an architecture and workflow that aids in increasing performance and faster deployments across infrastructures.

Architecture

Click the image to enlarge it.

Workflow

Click the image to enlarge it.

Supported Frameworks

Supported deep learning frameworks are:

• TensorFlow*, including 1.15.0 UP3, 2.7.0, 2.8.0, Official TensorFlow 2.6.2, Official TensorFlow 2.7.0, Official TensorFlow 2.8.0

Note: Intel Optimized TensorFlow 2.5.0 requires to set environment variable TF_ENABLE_MKL_NATIVE_FORMAT=0 before running Neural Compressor quantization or deploying the quantized model.

Note: From the official TensorFlow 2.6.0, oneDNN support has been upstreamed. Download the official TensorFlow 2.6.0 binary for the CPU device and set the environment variable TF_ENABLE_ONEDNN_OPTS=1 before running the quantization process or deploying the quantized model.

• PyTorch*, including 1.8.0+cpu, 1.9.0+cpu, 1.10.0+cpu
• Apache* MXNet, including 1.6.0, 1.7.0, 1.8.0
• ONNX* Runtime, including 1.8.0, 1.9.0, 1.10.0
• Execution Engine, a reference bare metal solution(./engine) for domain-specific NLP models.

Installation

Select the installation based on your operating system.

Linux Installation

You can install Neural Compressor using one of three options: Install just the library from binary or source, or get the Intel-optimized framework together with the library by installing the Intel® oneAPI AI Analytics Toolkit.

Prerequisites

The following prerequisites and requirements must be satisfied for a successful installation:

• Python version: 3.7 or 3.8 or 3.9

• C++ compiler: 7.2.1 or above

• CMake: 3.12 or above

common build issues

Issue 1: ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject

Solution: reinstall pycocotools by "pip install pycocotools --no-cache-dir"

Issue 2: ImportError: libGL.so.1: cannot open shared object file: No such file or directory

Solution: apt install or yum install opencv

Option 1 Install from binary

# install stable version from pip
pip install neural-compressor

# install nightly version from pip
pip install -i https://test.pypi.org/simple/ neural-compressor

# install stable version from from conda
conda install neural-compressor -c conda-forge -c intel 

Option 2 Install from source

git clone https://github.com/intel/neural-compressor.git
cd neural-compressor
git submodule sync
git submodule update --init --recursive
pip install -r requirements.txt
python setup.py install

Option 3 Install from AI Kit

The Intel® Neural Compressor library is released as part of the Intel® oneAPI AI Analytics Toolkit (AI Kit). The AI Kit provides a consolidated package of Intel's latest deep learning and machine optimizations all in one place for ease of development. Along with Neural Compressor, the AI Kit includes Intel-optimized versions of deep learning frameworks (such as TensorFlow and PyTorch) and high-performing Python libraries to streamline end-to-end data science and AI workflows on Intel architectures.

The AI Kit is distributed through many common channels, including from Intel's website, YUM, APT, Anaconda, and more. Select and download the AI Kit distribution package that's best suited for you and follow the Get Started Guide for post-installation instructions.

Download AI Kit	AI Kit Get Started Guide

Windows Installation

Prerequisites

The following prerequisites and requirements must be satisfied for a successful installation:

• Python version: 3.7 or 3.8 or 3.9

• Download and install anaconda.

• Create a virtual environment named nc in anaconda:

  # Here we install python 3.7 for instance. You can also choose python 3.8 or 3.9.
  conda create -n nc python=3.7
  conda activate nc

Installation options

Option 1 Install from binary

# install stable version from pip
pip install neural-compressor

# install nightly version from pip
pip install -i https://test.pypi.org/simple/ neural-compressor

# install from conda
conda install pycocotools -c esri   
conda install neural-compressor -c conda-forge -c intel

Option 2 Install from source

git clone https://github.com/intel/neural-compressor.git
cd neural-compressor
git submodule sync
git submodule update --init --recursive
pip install -r requirements.txt
python setup.py install

Documentation

Get Started

• APIs explains Intel® Neural Compressor's API.
• GUI provides web-based UI service to make quantization easier.
• Transform introduces how to utilize Neural Compressor's built-in data processing and how to develop a custom data processing method.
• Dataset introduces how to utilize Neural Compressor's built-in dataset and how to develop a custom dataset.
• Metric introduces how to utilize Neural Compressor's built-in metrics and how to develop a custom metric.
• Tutorial provides comprehensive instructions on how to utilize Neural Compressor's features with examples.
• Examples are provided to demonstrate the usage of Neural Compressor in different frameworks: TensorFlow, PyTorch, MXNet, and ONNX Runtime.
• Intel oneAPI AI Analytics Toolkit Get Started Guide explains the AI Kit components, installation and configuration guides, and instructions for building and running sample apps.
• AI and Analytics Samples includes code samples for Intel oneAPI libraries.

Deep Dive

• Quantization are processes that enable inference and training by performing computations at low-precision data types, such as fixed-point integers. Neural Compressor supports Post-Training Quantization (PTQ) with different quantization capabilities and Quantization-Aware Training (QAT). Note that (Dynamic Quantization) currently has limited support.
• Pruning provides a common method for introducing sparsity in weights and activations.
• Knowledge Distillation provides a common method for distilling knowledge from teacher model to student model.
• Distributed Training introduces how to leverage Horovod to do multi-node training in Intel® Neural Compressor to speed up the training time.
• Benchmarking introduces how to utilize the benchmark interface of Neural Compressor.
• Mixed precision introduces how to enable mixed precision, including BFP16 and int8 and FP32, on Intel platforms during tuning.
• Graph Optimization introduces how to enable graph optimization for FP32 and auto-mixed precision.
• Model Conversion introduces how to convert TensorFlow QAT model to quantized model running on Intel platforms.
• TensorBoard provides tensor histograms and execution graphs for tuning debugging purposes.

Advanced Topics

• Execution Engine is a bare metal solution domain-specific NLP models as the reference for customers.
• Adaptor is the interface between components and framework. The method to develop adaptor extension is introduced with ONNX Runtime as example.
• Strategy can automatically optimized low-precision recipes for deep learning models to achieve optimal product objectives like inference performance and memory usage with expected accuracy criteria. The method to develop a new strategy is introduced.

Publications

• New instructions in the Intel® Xeon® Scalable processors combined with optimized software frameworks enable real-time AI within network workloads (Feb 2022)
• Quantizing ONNX Models using Intel® Neural Compressor (Feb 2022)
• Quantize AI Model by Intel® oneAPI AI Analytics Toolkit on Alibaba Cloud (Feb 2022)
• AI Performance and Productivity with Intel® Neural Compressor (Jan 2022)
• Ease-of-use quantization for PyTorch with Intel® Neural Compressor (Jan 2022)
• Faster, Easier Optimization with Intel® Neural Compressor (Nov 2021)
• MLPerf™ Performance Gains Abound with latest 3rd Generation Intel® Xeon® Scalable Processors (Apr 2021)
• 3D Digital Face Reconstruction Solution enabled by 3rd Gen Intel® Xeon® Scalable Processors (Apr 2021)
• Accelerating Alibaba Transformer model performance with 3rd Gen Intel® Xeon® Scalable Processors (Ice Lake) and Intel® Deep Learning Boost (Apr 2021)
• Using Low-Precision Optimizations for High-Performance DL Inference Applications (Apr 2021)
• DL Boost Quantization with CERN's 3D-GANs model (Feb 2021)

Full publication list please refers to here

System Requirements

Intel® Neural Compressor supports systems based on Intel 64 architecture or compatible processors, specially optimized for the following CPUs:

• Intel Xeon Scalable processor (formerly Skylake, Cascade Lake, Cooper Lake, and Icelake)
• future Intel Xeon Scalable processor (code name Sapphire Rapids)

Intel® Neural Compressor requires installing the Intel-optimized framework version for the supported DL framework you use: TensorFlow, PyTorch, MXNet, or ONNX runtime.

Note: Intel Neural Compressor supports Intel-optimized and official frameworks for some TensorFlow versions. Refer to Supported Frameworks for specifics.

Validated Hardware/Software Environment

Platform	OS	Python	Framework	Version
Cascade Lake Cooper Lake Skylake Ice Lake	CentOS 8.3 Ubuntu 18.04	3.7 3.8 3.9	TensorFlow	2.8.0
				2.7.0
				2.6.2
				1.15.0UP3
			PyTorch	1.10.0+cpu
				1.9.0+cpu
				1.8.0+cpu
				IPEX
			MXNet	1.8.0
				1.7.0
				1.6.0
			ONNX Runtime	1.10.0
				1.9.0
				1.8.0

Validated Models

Intel® Neural Compressor provides numerous examples to show promising accuracy loss with the best performance gain. A full quantized model list on various frameworks is available in the Model List.

Validated MLPerf Models

Model	Framework	Support	Example
ResNet50 v1.5	TensorFlow	Yes	Link
	PyTorch	Yes	Link
DLRM	PyTorch	Yes	Link
BERT-large	TensorFlow	Yes	Link
	PyTorch	Yes	Link
SSD-ResNet34	TensorFlow	Yes	Link
	PyTorch	Yes	Link
RNN-T	PyTorch	Yes	Link
3D-UNet	TensorFlow	WIP
	PyTorch	Yes	Link

Validated Quantized Models

Framework	version	model	Accuracy			Performance/ICX8380/1s4c10ins1bs/throughput(samples/sec)
			INT8	FP32	Acc Ratio[(INT8-FP32)/FP32]	INT8	FP32	Performance Ratio[INT8/FP32]
tensorflow	2.6.0	resnet50v1.0	74.11%	74.27%	-0.22%	1287.00	495.29	2.60x
tensorflow	2.6.0	resnet50v1.5	76.82%	76.46%	0.47%	1218.03	420.34	2.90x
tensorflow	2.6.0	resnet101	77.50%	76.45%	1.37%	849.62	345.54	2.46x
tensorflow	2.6.0	inception_v1	70.48%	69.74%	1.06%	2202.64	1058.20	2.08x
tensorflow	2.6.0	inception_v2	74.36%	73.97%	0.53%	1751.31	827.81	2.11x
tensorflow	2.6.0	inception_v3	77.28%	76.75%	0.69%	868.06	384.17	2.26x
tensorflow	2.6.0	inception_v4	80.40%	80.27%	0.16%	569.48	197.28	2.89x
tensorflow	2.6.0	inception_resnet_v2	80.44%	80.40%	0.05%	269.03	137.25	1.96x
tensorflow	2.6.0	mobilenetv1	71.79%	70.96%	1.17%	3831.42	1189.06	3.22x
tensorflow	2.6.0	mobilenetv2	71.79%	71.76%	0.04%	2570.69	1237.62	2.07x
tensorflow	2.6.0	ssd_resnet50_v1	37.86%	38.00%	-0.37%	65.52	24.01	2.73x
tensorflow	2.6.0	ssd_mobilenet_v1	22.97%	23.13%	-0.69%	842.46	404.04	2.08x
tensorflow	2.6.0	ssd_resnet34	21.69%	22.09%	-1.81%	41.23	10.75	3.83x

Framework	version	model	Accuracy			Performance/ICX8380/1s4c10ins1bs/throughput(samples/sec)
			INT8	FP32	Acc Ratio[(INT8-FP32)/FP32]	INT8	FP32	Performance Ratio[INT8/FP32]
pytorch	1.9.0+cpu	resnet18	69.59%	69.76%	-0.24%	692.04	363.64	1.90x
pytorch	1.9.0+cpu	resnet50	76.00%	76.13%	-0.17%	453.10	186.67	2.43x
pytorch	1.9.0+cpu	resnext101_32x8d	79.02%	79.31%	-0.36%	196.27	70.08	2.80x
pytorch	1.9.0+cpu	bert_base_mrpc	88.12%	88.73%	-0.69%	199.32	107.34	1.86x
pytorch	1.9.0+cpu	bert_base_cola	59.06%	58.84%	0.37%	198.53	105.29	1.89x
pytorch	1.9.0+cpu	bert_base_sts-b	88.72%	89.27%	-0.62%	203.29	107.03	1.90x
pytorch	1.9.0+cpu	bert_base_sst-2	91.74%	91.86%	-0.13%	197.86	105.31	1.88x
pytorch	1.9.0+cpu	bert_base_rte	70.40%	69.68%	1.04%	192.90	107.25	1.80x
pytorch	1.9.0+cpu	bert_large_mrpc	87.66%	88.33%	-0.75%	94.08	33.84	2.78x
pytorch	1.9.0+cpu	bert_large_squad	92.69	93.05	-0.38%	20.93	11.18	1.87x
pytorch	1.9.0+cpu	bert_large_qnli	91.12%	91.82%	-0.76%	93.75	33.73	2.78x
pytorch	1.9.0+cpu	bert_large_rte	72.20%	72.56%	-0.50%	52.80	33.62	1.57x
pytorch	1.9.0+cpu	bert_large_cola	62.07%	62.57%	-0.80%	94.97	33.77	2.81x
pytorch	1.9.0+cpu	inception_v3	69.48%	69.54%	-0.09%	418.59	207.77	2.01x
pytorch	1.9.0+cpu	peleenet	71.61%	72.08%	-0.66%	461.47	359.58	1.28x
pytorch	1.9.0+cpu	yolo_v3	24.50%	24.54%	-0.17%	98.11	37.50	2.62x

Validated Pruning Models

Tasks	FWK	Model	fp32 baseline	gradient sensitivity with 20% sparsity			+onnx dynamic quantization on pruned model
				accuracy%	drop%	perf gain (sample/s)	accuracy%	drop%	perf gain (sample/s)
SST-2	pytorch	bert-base	accuracy = 92.32	accuracy = 91.97	-0.38	1.30x	accuracy = 92.20	-0.13	1.86x
QQP	pytorch	bert-base	[accuracy, f1] = [91.10, 88.05]	[accuracy, f1] = [89.97, 86.54]	[-1.24, -1.71]	1.32x	[accuracy, f1] = [89.75, 86.60]	[-1.48, -1.65]	1.81x

Tasks	FWK	Model	fp32 baseline	Pattern Lock on 70% Unstructured Sparsity		Pattern Lock on 50% 1:2 Structured Sparsity
				accuracy%	drop%	accuracy%	drop%
MNLI	pytorch	bert-base	[m, mm] = [84.57, 84.79]	[m, mm] = [82.45, 83.27]	[-2.51, -1.80]	[m, mm] = [83.20, 84.11]	[-1.62, -0.80]
SST-2	pytorch	bert-base	accuracy = 92.32	accuracy = 91.51	-0.88	accuracy = 92.20	-0.13
QQP	pytorch	bert-base	[accuracy, f1] = [91.10, 88.05]	[accuracy, f1] = [90.48, 87.06]	[-0.68, -1.12]	[accuracy, f1] = [90.92, 87.78]	[-0.20, -0.31]
QNLI	pytorch	bert-base	accuracy = 91.54	accuracy = 90.39	-1.26	accuracy = 90.87	-0.73
QnA	pytorch	bert-base	[em, f1] = [79.34, 87.10]	[em, f1] = [77.27, 85.75]	[-2.61, -1.54]	[em, f1] = [78.03, 86.50]	[-1.65, -0.69]

Framework	Model	fp32 baseline	Compression	dataset	acc(drop)%
Pytorch	resnet18	69.76	30% sparsity on magnitude	ImageNet	69.47(-0.42)
Pytorch	resnet18	69.76	30% sparsity on gradient sensitivity	ImageNet	68.85(-1.30)
Pytorch	resnet50	76.13	30% sparsity on magnitude	ImageNet	76.11(-0.03)
Pytorch	resnet50	76.13	30% sparsity on magnitude and post training quantization	ImageNet	76.01(-0.16)
Pytorch	resnet50	76.13	30% sparsity on magnitude and quantization aware training	ImageNet	75.90(-0.30)

Validated Knowledge Distillation Examples

Example Name	Dataset	Student (Accuracy)	Teacher (Accuracy)	Student With Distillation (Accuracy Improvement)
ResNet example	ImageNet	ResNet18 (0.6739)	ResNet50 (0.7399)	0.6845 (0.0106)
BlendCnn example	MRPC	BlendCnn (0.7034)	BERT-Base (0.8382)	0.7034 (0)
BiLSTM example	SST-2	BiLSTM (0.7913)	RoBERTa-Base (0.9404)	0.8085 (0.0172)

Validated Engine Examples

model	Accuracy			Performance/ICX8380/1s4c10ins1bs/seq_len128/throughput(samples/sec)			Performance/ICX8380/2s4c20ins64bs/seq_len128/throughput(samples/sec)
	INT8	FP32	Acc   Ratio[(INT8-FP32)/FP32]	INT8	FP32	Preformance   Ratio[INT8/FP32]	INT8	FP32	Preformance   Ratio[INT8/FP32]
bert_large_squad	90.74	90.87	-0.14%	44.9	12.33	3.64x	362.21	88.38	4.10x
distilbert_base_uncased_sst2	90.14%	90.25%	-0.12%	1003.01	283.69	3.54x	2104.26	606.58	3.47x
minilm_l6_h384_uncased_sst2	89.33%	90.14%	-0.90%	2739.73	999	2.74x	5389.98	2333.14	2.31x
roberta_base_mrpc	89.46%	88.97%	0.55%	506.07	142.13	3.56x	1167.09	311.5	3.75x
bert_base_nli_mean_tokens_stsb	89.27%	89.55%	-0.31%	503.52	140.98	3.57x	1096.46	332.54	3.30x
bert_base_sparse_mrpc	70.34%	70.59%	-0.35%	506.59	142.33	3.56x	1133.04	339.96	3.33x
distilroberta_base_wnli	56.34%	56.34%	0.00%	1026.69	290.7	3.53x	2309.9	620.81	3.72x
paraphrase_xlm_r_multilingual_v1_stsb	86.72%	87.23%	-0.58%	509.68	142.73	3.57x	1169.45	311.59	3.75x
distilbert_base_uncased_mrpc	84.07%	84.07%	0.00%	1002	280.27	3.58x	2107.96	606.95	3.47x
finbert_financial_phrasebank	82.74%	82.80%	-0.07%	919.12	272.48	3.37x	1101.13	331.88	3.32x
distilbert_base_uncased_emotion	93.85%	94.20%	-0.37%	1003.01	283.53	3.54x	2103.22	607.08	3.46x

Additional Content

• Release Information
• Contribution Guidelines
• Legal
• Security Policy
• Intel® Neural Compressor Website

Hiring

We are hiring. Please send your resume to inc.maintainers@intel.com if you have interests in model compression techniques.