dlquantizer

Quantize a deep neural network to 8-bit scaled integer data types

Since R2020a

Description

Use the dlquantizer object to reduce the memory requirement of a deep neural network by quantizing weights, biases, and activations to 8-bit scaled integer data types. You can create and verify the behavior of a quantized network for GPU, FPGA, CPU deployment, or explore the quantized network in MATLAB^®.

For CPU and GPU deployment, the software generates code for a convolutional deep neural network by quantizing the weights, biases, and activations of the convolution layers to 8-bit scaled integer data types. The quantization is performed by providing the calibration result file produced by the calibrate function to the codegen (MATLAB Coder) command. Note that code generation does not support quantized deep neural networks produced by the quantize function.

This object requires Deep Learning Toolbox Model Quantization Library. To learn about the products required to quantize a deep neural network, see Quantization Workflow Prerequisites.

Creation

Syntax

quantObj = dlquantizer(net)

quantObj = dlquantizer(net,Name,Value)

Description

quantObj = dlquantizer(net) creates a dlquantizer object for the specified deep neural network, net.

quantObj = dlquantizer(net,Name,Value) creates a dlquantizer object for the specified network, with additional options specified by one or more name-value pair arguments.

Input Arguments

net — Pretrained neural network

DAGNetwork object | dlnetwork object | SeriesNetwork object | yolov2ObjectDetector object | yolov3ObjectDetector object | yolov4ObjectDetector object | ssdObjectDetector object

Pretrained neural network, specified as a DAGNetwork, dlnetwork, SeriesNetwork, yolov2ObjectDetector (Computer Vision Toolbox), yolov3ObjectDetector (Computer Vision Toolbox), yolov4ObjectDetector (Computer Vision Toolbox), or a ssdObjectDetector (Computer Vision Toolbox) object.

Properties

NetworkObject — Pretrained neural network

DAGNetwork object | dlnetwork object | SeriesNetwork object | yolov2ObjectDetector object | yolov3ObjectDetector | yolov4ObjectDetector | ssdObjectDetector object

This property is read-only.

Pre-trained neural network, specified as a DAGNetwork, dlnetwork, SeriesNetwork, yolov2ObjectDetector (Computer Vision Toolbox), yolov3ObjectDetector (Computer Vision Toolbox), yolov4ObjectDetector (Computer Vision Toolbox), or a ssdObjectDetector (Computer Vision Toolbox) object.

ExecutionEnvironment — Execution environment

'GPU' (default) | 'FPGA' | 'CPU' | 'MATLAB'

Execution environment for the quantized network, specified as 'GPU', 'FPGA', 'CPU', or 'MATLAB'. How the network is quantized depends on the choice of execution environment.

The 'MATLAB' execution environment indicates a target-agnostic quantization of the neural network will be performed. This option does not require you to have target hardware in order to explore the quantized network in MATLAB.

Example: 'ExecutionEnvironment','FPGA'

Object Functions

calibrate	Simulate and collect ranges of a deep neural network
validate	Quantize and validate a deep neural network
quantize	Quantize deep neural network

Examples

Quantize a Neural Network for GPU Target

This example shows how to quantize learnable parameters in the convolution layers of a neural network for GPU and explore the behavior of the quantized network. In this example, you quantize the squeezenet neural network after retraining the network to classify new images according to the Train Deep Learning Network to Classify New Images example. In this example, the memory required for the network is reduced approximately 75% through quantization while the accuracy of the network is not affected.

Load the pretrained network. net is the output network of the Train Deep Learning Network to Classify New Images example.

load squeezenetmerch
net

net = 
  DAGNetwork with properties:

         Layers: [68×1 nnet.cnn.layer.Layer]
    Connections: [75×2 table]
     InputNames: {'data'}
    OutputNames: {'new_classoutput'}

Define calibration and validation data to use for quantization.

The calibration data is used to collect the dynamic ranges of the weights and biases in the convolution and fully connected layers of the network and the dynamic ranges of the activations in all layers of the network. For the best quantization results, the calibration data must be representative of inputs to the network.

The validation data is used to test the network after quantization to understand the effects of the limited range and precision of the quantized convolution layers in the network.

In this example, use the images in the MerchData data set. Define an augmentedImageDatastore object to resize the data for the network. Then, split the data into calibration and validation data sets.

unzip('MerchData.zip');
imds = imageDatastore('MerchData', ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');
[calData, valData] = splitEachLabel(imds, 0.7, 'randomized');
aug_calData = augmentedImageDatastore([227 227], calData);
aug_valData = augmentedImageDatastore([227 227], valData);

Create a dlquantizer object and specify the network to quantize.

dlquantObj = dlquantizer(net);

Specify the GPU target.

quantOpts = dlquantizationOptions(Target,'gpu');

Use the calibrate function to exercise the network with sample inputs and collect range information. The calibrate function exercises the network and collects the dynamic ranges of the weights and biases in the convolution and fully connected layers of the network and the dynamic ranges of the activations in all layers of the network. The function returns a table. Each row of the table contains range information for a learnable parameter of the optimized network.

calResults = calibrate(dlquantObj, aug_calData)

calResults=121×5 table
        Optimized Layer Name         Network Layer Name     Learnables / Activations    MinValue     MaxValue
    ____________________________    ____________________    ________________________    _________    ________

    {'conv1_Weights'           }    {'conv1'           }           "Weights"             -0.91985     0.88489
    {'conv1_Bias'              }    {'conv1'           }           "Bias"                -0.07925     0.26343
    {'fire2-squeeze1x1_Weights'}    {'fire2-squeeze1x1'}           "Weights"                -1.38      1.2477
    {'fire2-squeeze1x1_Bias'   }    {'fire2-squeeze1x1'}           "Bias"                -0.11641     0.24273
    {'fire2-expand1x1_Weights' }    {'fire2-expand1x1' }           "Weights"              -0.7406     0.90982
    {'fire2-expand1x1_Bias'    }    {'fire2-expand1x1' }           "Bias"               -0.060056     0.14602
    {'fire2-expand3x3_Weights' }    {'fire2-expand3x3' }           "Weights"             -0.74397     0.66905
    {'fire2-expand3x3_Bias'    }    {'fire2-expand3x3' }           "Bias"               -0.051778    0.074239
    {'fire3-squeeze1x1_Weights'}    {'fire3-squeeze1x1'}           "Weights"              -0.7712     0.68917
    {'fire3-squeeze1x1_Bias'   }    {'fire3-squeeze1x1'}           "Bias"                -0.10138     0.32675
    {'fire3-expand1x1_Weights' }    {'fire3-expand1x1' }           "Weights"             -0.72035      0.9743
    {'fire3-expand1x1_Bias'    }    {'fire3-expand1x1' }           "Bias"               -0.067029     0.30425
    {'fire3-expand3x3_Weights' }    {'fire3-expand3x3' }           "Weights"             -0.61443      0.7741
    {'fire3-expand3x3_Bias'    }    {'fire3-expand3x3' }           "Bias"               -0.053613     0.10329
    {'fire4-squeeze1x1_Weights'}    {'fire4-squeeze1x1'}           "Weights"              -0.7422      1.0877
    {'fire4-squeeze1x1_Bias'   }    {'fire4-squeeze1x1'}           "Bias"                -0.10885     0.13881
      ⋮

Use the validate function to quantize the learnable parameters in the convolution layers of the network and exercise the network. The function uses the metric function defined in the dlquantizationOptions object to compare the results of the network before and after quantization.

valResults = validate(dlquantObj, aug_valData, quantOpts)

valResults = struct with fields:
       NumSamples: 20
    MetricResults: [1×1 struct]
       Statistics: [2×2 table]

Examine the validation output to see the performance of the quantized network.

valResults.MetricResults.Result

ans=2×2 table
    NetworkImplementation    MetricOutput
    _____________________    ____________

     {'Floating-Point'}           1      
     {'Quantized'     }           1      

valResults.Statistics

ans=2×2 table
    NetworkImplementation    LearnableParameterMemory(bytes)
    _____________________    _______________________________

     {'Floating-Point'}                2.9003e+06           
     {'Quantized'     }                7.3393e+05           

In this example, the memory required for the network was reduced approximately 75% through quantization. The accuracy of the network is not affected.

The weights, biases, and activations of the convolution layers of the network specified in the dlquantizer object now use scaled 8-bit integer data types.

Quantize a Neural Network for CPU Target

This example shows how to quantize and validate a neural network for a CPU target. This workflow is similar to other execution environments, but before validating you must establish a raspi connection and specify it as target using dlquantizationOptions.

First, load your network. This example uses the pretrained network squeezenet.

load squeezenetmerch
net

net = 
  DAGNetwork with properties:

         Layers: [68×1 nnet.cnn.layer.Layer]
    Connections: [75×2 table]
     InputNames: {'data'}
    OutputNames: {'new_classoutput'}

Then define your calibration and validation data, calDS and valDS respectively.

unzip('MerchData.zip');
imds = imageDatastore('MerchData', ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');
[calData, valData] = splitEachLabel(imds, 0.7, 'randomized');
aug_calData = augmentedImageDatastore([227 227],calData);
aug_valData = augmentedImageDatastore([227 227],valData);

Create the dlquantizer object and specify a CPU execution environment.

dq =  dlquantizer(net,'ExecutionEnvironment','CPU') 

dq = 
  dlquantizer with properties:

           NetworkObject: [1×1 DAGNetwork]
    ExecutionEnvironment: 'CPU'

Calibrate the network.

calResults = calibrate(dq,aug_calData,'UseGPU','off')

calResults=122×5 table
        Optimized Layer Name         Network Layer Name     Learnables / Activations    MinValue     MaxValue
    ____________________________    ____________________    ________________________    _________    ________

    {'conv1_Weights'           }    {'conv1'           }           "Weights"             -0.91985     0.88489
    {'conv1_Bias'              }    {'conv1'           }           "Bias"                -0.07925     0.26343
    {'fire2-squeeze1x1_Weights'}    {'fire2-squeeze1x1'}           "Weights"                -1.38      1.2477
    {'fire2-squeeze1x1_Bias'   }    {'fire2-squeeze1x1'}           "Bias"                -0.11641     0.24273
    {'fire2-expand1x1_Weights' }    {'fire2-expand1x1' }           "Weights"              -0.7406     0.90982
    {'fire2-expand1x1_Bias'    }    {'fire2-expand1x1' }           "Bias"               -0.060056     0.14602
    {'fire2-expand3x3_Weights' }    {'fire2-expand3x3' }           "Weights"             -0.74397     0.66905
    {'fire2-expand3x3_Bias'    }    {'fire2-expand3x3' }           "Bias"               -0.051778    0.074239
    {'fire3-squeeze1x1_Weights'}    {'fire3-squeeze1x1'}           "Weights"              -0.7712     0.68917
    {'fire3-squeeze1x1_Bias'   }    {'fire3-squeeze1x1'}           "Bias"                -0.10138     0.32675
    {'fire3-expand1x1_Weights' }    {'fire3-expand1x1' }           "Weights"             -0.72035      0.9743
    {'fire3-expand1x1_Bias'    }    {'fire3-expand1x1' }           "Bias"               -0.067029     0.30425
    {'fire3-expand3x3_Weights' }    {'fire3-expand3x3' }           "Weights"             -0.61443      0.7741
    {'fire3-expand3x3_Bias'    }    {'fire3-expand3x3' }           "Bias"               -0.053613     0.10329
    {'fire4-squeeze1x1_Weights'}    {'fire4-squeeze1x1'}           "Weights"              -0.7422      1.0877
    {'fire4-squeeze1x1_Bias'   }    {'fire4-squeeze1x1'}           "Bias"                -0.10885     0.13881
      ⋮

Use the MATLAB Support Package for Raspberry Pi Hardware function, raspi, to create a connection to the Raspberry Pi. In the following code, replace:

• raspiname with the name or address of your Raspberry Pi

• username with your user name

• password with your password

% r = raspi('raspiname','username','password')

For example,

r = raspi('gpucoder-raspberrypi-7','pi','matlab')

r = 
  raspi with properties:

         DeviceAddress: 'gpucoder-raspberrypi-7'      
                  Port: 18734                         
             BoardName: 'Raspberry Pi 3 Model B+'     
         AvailableLEDs: {'led0'}                      
  AvailableDigitalPins: [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27]
  AvailableSPIChannels: {}                            
     AvailableI2CBuses: {}                            
      AvailableWebcams: {}                            
           I2CBusSpeed:                               
AvailableCANInterfaces: {}                            

  Supported peripherals

Specify raspi object as the target for the quantized network.

opts = dlquantizationOptions('Target',r)

opts = 
  dlquantizationOptions with properties:

    MetricFcn: {}
    Bitstream: ''
       Target: [1×1 raspi]

Validate the quantized network with the validate function.

valResults = validate(dq,aug_valData,opts)

### Starting application: 'codegen\lib\validate_predict_int8\pil\validate_predict_int8.elf'
    To terminate execution: clear validate_predict_int8_pil
### Launching application validate_predict_int8.elf...
### Host application produced the following standard output (stdout) and standard error (stderr) messages:

valResults = struct with fields:
       NumSamples: 20
    MetricResults: [1×1 struct]
       Statistics: []

Examine the validation output to see the performance of the quantized network.

valResults.MetricResults.Result

ans=2×2 table
    NetworkImplementation    MetricOutput
    _____________________    ____________

     {'Floating-Point'}          0.95    
     {'Quantized'     }          0.95    

Quantize Network for FPGA Deployment

Reduce the memory footprint of a deep neural network by quantizing the weights, biases, and activations of convolution layers to 8-bit scaled integer data types. This example shows how to use Deep Learning Toolbox Model Quantization Library and Deep Learning HDL Toolbox to deploy the int8 network to a target FPGA board.

For this example, you need:

• Deep Learning Toolbox™

• Deep Learning HDL Toolbox™

• Deep Learning Toolbox Model Quantization Library

• Deep Learning HDL Toolbox Support Package for Xilinx® FPGA and SoC Devices

• MATLAB Coder Interface for Deep Learning.

Load Pretrained Network

Load the pretrained LogoNet network and analyze the network architecture.

snet = getLogoNetwork;
deepNetworkDesigner(snet);

Set random number generator for reproducibility.

rng(0);

Load Data

This example uses the logos_dataset data set. The data set consists of 320 images. Each image is 227-by-227 in size and has three color channels (RGB). Create an augmentedImageDatastore object for calibration and validation.

curDir = pwd;
unzip("logos_dataset.zip");
imageData = imageDatastore(fullfile(curDir,'logos_dataset'),...
'IncludeSubfolders',true,'FileExtensions','.JPG','LabelSource','foldernames');
[calibrationData, validationData] = splitEachLabel(imageData, 0.5,'randomized');

Generate Calibration Result File for the Network

Create a dlquantizer (Deep Learning HDL Toolbox) object and specify the network to quantize. Specify the execution environment as FPGA.

dlQuantObj = dlquantizer(snet,'ExecutionEnvironment',"FPGA");

Use the calibrate (Deep Learning HDL Toolbox) function to exercise the network with sample inputs and collect the range information. The calibrate function collects the dynamic ranges of the weights and biases. The calibrate function returns a table. Each row of the table contains range information for a learnable parameter of the quantized network.

calibrate(dlQuantObj,calibrationData)

ans=35×5 table
        Optimized Layer Name        Network Layer Name    Learnables / Activations     MinValue       MaxValue 
    ____________________________    __________________    ________________________    ___________    __________

    {'conv_1_Weights'          }      {'conv_1'    }           "Weights"                -0.048978      0.039352
    {'conv_1_Bias'             }      {'conv_1'    }           "Bias"                     0.99996        1.0028
    {'conv_2_Weights'          }      {'conv_2'    }           "Weights"                -0.055518      0.061901
    {'conv_2_Bias'             }      {'conv_2'    }           "Bias"                 -0.00061171       0.00227
    {'conv_3_Weights'          }      {'conv_3'    }           "Weights"                -0.045942      0.046927
    {'conv_3_Bias'             }      {'conv_3'    }           "Bias"                  -0.0013998     0.0015218
    {'conv_4_Weights'          }      {'conv_4'    }           "Weights"                -0.045967         0.051
    {'conv_4_Bias'             }      {'conv_4'    }           "Bias"                    -0.00164     0.0037892
    {'fc_1_Weights'            }      {'fc_1'      }           "Weights"                -0.051394      0.054344
    {'fc_1_Bias'               }      {'fc_1'      }           "Bias"                 -0.00052319    0.00084454
    {'fc_2_Weights'            }      {'fc_2'      }           "Weights"                 -0.05016      0.051557
    {'fc_2_Bias'               }      {'fc_2'      }           "Bias"                  -0.0017564     0.0018502
    {'fc_3_Weights'            }      {'fc_3'      }           "Weights"                -0.050706       0.04678
    {'fc_3_Bias'               }      {'fc_3'      }           "Bias"                    -0.02951      0.024855
    {'imageinput'              }      {'imageinput'}           "Activations"                    0           255
    {'imageinput_normalization'}      {'imageinput'}           "Activations"              -139.34        198.72
      ⋮

Create Target Object

Create a target object with a custom name for your target device and an interface to connect your target device to the host computer. Interface options are JTAG and Ethernet. Interface options are JTAG and Ethernet. To use JTAG, install Xilinx Vivado® Design Suite 2022.1. To set the Xilinx Vivado toolpath, enter:

hdlsetuptoolpath('ToolName', 'Xilinx Vivado', 'ToolPath', 'C:\Xilinx\Vivado\2022.1\bin\vivado.bat');

To create the target object, enter:

hTarget = dlhdl.Target('Xilinx','Interface','Ethernet','IPAddress','10.10.10.15');

Alternatively, you can also use the JTAG interface.

% hTarget = dlhdl.Target('Xilinx', 'Interface', 'JTAG');

Create dlQuantizationOptions Object

Create a dlquantizationOptions object. Specify the target bitstream and target board interface. The default metric function is a Top-1 accuracy metric function.

options_FPGA = dlquantizationOptions('Bitstream','zcu102_int8','Target',hTarget);
options_emulation = dlquantizationOptions('Target','host');

To use a custom metric function, specify the metric function in the dlquantizationOptions object.

options_FPGA = dlquantizationOptions('MetricFcn',{@(x)hComputeAccuracy(x,snet,validationData)},'Bitstream','zcu102_int8','Target',hTarget);
options_emulation = dlquantizationOptions('MetricFcn',{@(x)hComputeAccuracy(x,snet,validationData)})

Validate Quantized Network

Use the validate function to quantize the learnable parameters in the convolution layers of the network. The validate function simulates the quantized network in MATLAB. The validate function uses the metric function defined in the dlquantizationOptions object to compare the results of the single-data-type network object to the results of the quantized network object.

prediction_emulation = dlQuantObj.validate(validationData,options_emulation)

prediction_emulation = struct with fields:
       NumSamples: 160
    MetricResults: [1×1 struct]
       Statistics: []

For validation on an FPGA, the validate function:

• Programs the FPGA board by using the output of the compile method and the programming file

• Downloads the network weights and biases

• Compares the performance of the network before and after quantization

prediction_FPGA = dlQuantObj.validate(validationData,options_FPGA)

### Compiling network for Deep Learning FPGA prototyping ...
### Targeting FPGA bitstream zcu102_int8.
### The network includes the following layers:
     1   'imageinput'    Image Input             227×227×3 images with 'zerocenter' normalization and 'randfliplr' augmentations  (SW Layer)
     2   'conv_1'        2-D Convolution         96 5×5×3 convolutions with stride [1  1] and padding [0  0  0  0]                (HW Layer)
     3   'relu_1'        ReLU                    ReLU                                                                             (HW Layer)
     4   'maxpool_1'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
     5   'conv_2'        2-D Convolution         128 3×3×96 convolutions with stride [1  1] and padding [0  0  0  0]              (HW Layer)
     6   'relu_2'        ReLU                    ReLU                                                                             (HW Layer)
     7   'maxpool_2'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
     8   'conv_3'        2-D Convolution         384 3×3×128 convolutions with stride [1  1] and padding [0  0  0  0]             (HW Layer)
     9   'relu_3'        ReLU                    ReLU                                                                             (HW Layer)
    10   'maxpool_3'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
    11   'conv_4'        2-D Convolution         128 3×3×384 convolutions with stride [2  2] and padding [0  0  0  0]             (HW Layer)
    12   'relu_4'        ReLU                    ReLU                                                                             (HW Layer)
    13   'maxpool_4'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
    14   'fc_1'          Fully Connected         2048 fully connected layer                                                       (HW Layer)
    15   'relu_5'        ReLU                    ReLU                                                                             (HW Layer)
    16   'fc_2'          Fully Connected         2048 fully connected layer                                                       (HW Layer)
    17   'relu_6'        ReLU                    ReLU                                                                             (HW Layer)
    18   'fc_3'          Fully Connected         32 fully connected layer                                                         (HW Layer)
    19   'softmax'       Softmax                 softmax                                                                          (SW Layer)
    20   'classoutput'   Classification Output   crossentropyex with 'adidas' and 31 other classes                                (SW Layer)
                                                                                                                                
### Notice: The layer 'imageinput' with type 'nnet.cnn.layer.ImageInputLayer' is implemented in software.
### Notice: The layer 'softmax' with type 'nnet.cnn.layer.SoftmaxLayer' is implemented in software.
### Notice: The layer 'classoutput' with type 'nnet.cnn.layer.ClassificationOutputLayer' is implemented in software.
### Compiling layer group: conv_1>>relu_4 ...
### Compiling layer group: conv_1>>relu_4 ... complete.
### Compiling layer group: maxpool_4 ...
### Compiling layer group: maxpool_4 ... complete.
### Compiling layer group: fc_1>>fc_3 ...
### Compiling layer group: fc_1>>fc_3 ... complete.

### Allocating external memory buffers:

          offset_name          offset_address    allocated_space 
    _______________________    ______________    ________________

    "InputDataOffset"           "0x00000000"     "11.9 MB"       
    "OutputResultOffset"        "0x00be0000"     "128.0 kB"      
    "SchedulerDataOffset"       "0x00c00000"     "128.0 kB"      
    "SystemBufferOffset"        "0x00c20000"     "9.9 MB"        
    "InstructionDataOffset"     "0x01600000"     "4.6 MB"        
    "ConvWeightDataOffset"      "0x01aa0000"     "8.2 MB"        
    "FCWeightDataOffset"        "0x022e0000"     "10.4 MB"       
    "EndOffset"                 "0x02d40000"     "Total: 45.2 MB"

### Network compilation complete.

### FPGA bitstream programming has been skipped as the same bitstream is already loaded on the target FPGA.
### Deep learning network programming has been skipped as the same network is already loaded on the target FPGA.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Finished writing input activations.
### Running single input activation.
### Notice: The layer 'imageinput' of type 'ImageInputLayer' is split into an image input layer 'imageinput' and an addition layer 'imageinput_norm' for normalization on hardware.
### The network includes the following layers:
     1   'imageinput'    Image Input             227×227×3 images with 'zerocenter' normalization and 'randfliplr' augmentations  (SW Layer)
     2   'conv_1'        2-D Convolution         96 5×5×3 convolutions with stride [1  1] and padding [0  0  0  0]                (HW Layer)
     3   'relu_1'        ReLU                    ReLU                                                                             (HW Layer)
     4   'maxpool_1'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
     5   'conv_2'        2-D Convolution         128 3×3×96 convolutions with stride [1  1] and padding [0  0  0  0]              (HW Layer)
     6   'relu_2'        ReLU                    ReLU                                                                             (HW Layer)
     7   'maxpool_2'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
     8   'conv_3'        2-D Convolution         384 3×3×128 convolutions with stride [1  1] and padding [0  0  0  0]             (HW Layer)
     9   'relu_3'        ReLU                    ReLU                                                                             (HW Layer)
    10   'maxpool_3'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
    11   'conv_4'        2-D Convolution         128 3×3×384 convolutions with stride [2  2] and padding [0  0  0  0]             (HW Layer)
    12   'relu_4'        ReLU                    ReLU                                                                             (HW Layer)
    13   'maxpool_4'     2-D Max Pooling         3×3 max pooling with stride [2  2] and padding [0  0  0  0]                      (HW Layer)
    14   'fc_1'          Fully Connected         2048 fully connected layer                                                       (HW Layer)
    15   'relu_5'        ReLU                    ReLU                                                                             (HW Layer)
    16   'fc_2'          Fully Connected         2048 fully connected layer                                                       (HW Layer)
    17   'relu_6'        ReLU                    ReLU                                                                             (HW Layer)
    18   'fc_3'          Fully Connected         32 fully connected layer                                                         (HW Layer)
    19   'softmax'       Softmax                 softmax                                                                          (SW Layer)
    20   'classoutput'   Classification Output   crossentropyex with 'adidas' and 31 other classes                                (SW Layer)
                                                                                                                                
### Notice: The layer 'softmax' with type 'nnet.cnn.layer.SoftmaxLayer' is implemented in software.
### Notice: The layer 'classoutput' with type 'nnet.cnn.layer.ClassificationOutputLayer' is implemented in software.


              Deep Learning Processor Estimator Performance Results

                   LastFrameLatency(cycles)   LastFrameLatency(seconds)       FramesNum      Total Latency     Frames/s
                         -------------             -------------              ---------        ---------       ---------
Network                   39136574                  0.17789                       1           39136574              5.6
    imageinput_norm         216472                  0.00098 
    conv_1                 6832680                  0.03106 
    maxpool_1              3705912                  0.01685 
    conv_2                10454501                  0.04752 
    maxpool_2              1173810                  0.00534 
    conv_3                 9364533                  0.04257 
    maxpool_3              1229970                  0.00559 
    conv_4                 1759348                  0.00800 
    maxpool_4                24450                  0.00011 
    fc_1                   2651288                  0.01205 
    fc_2                   1696632                  0.00771 
    fc_3                     26978                  0.00012 
 * The clock frequency of the DL processor is: 220MHz


### Finished writing input activations.
### Running single input activation.

prediction_FPGA = struct with fields:
       NumSamples: 160
    MetricResults: [1×1 struct]
       Statistics: [2×7 table]

View Performance of Quantized Neural Network

Display the accuracy of the quantized network.

prediction_emulation.MetricResults.Result

ans=2×2 table
    NetworkImplementation    MetricOutput
    _____________________    ____________

     {'Floating-Point'}         0.9875   
     {'Quantized'     }         0.9875   

prediction_FPGA.MetricResults.Result

ans=2×2 table
    NetworkImplementation    MetricOutput
    _____________________    ____________

     {'Floating-Point'}         0.9875   
     {'Quantized'     }         0.9875   

Display the performance of the quantized network in frames per second.

prediction_FPGA.Statistics

ans=2×7 table
    NetworkImplementation    FramesPerSecond    Number of Threads (Convolution)    Number of Threads (Fully Connected)    LUT Utilization (%)    BlockRAM Utilization (%)    DSP Utilization (%)
    _____________________    _______________    _______________________________    ___________________________________    ___________________    ________________________    ___________________

     {'Floating-Point'}          5.6213                       16                                    4                           93.198                    63.925                   15.595       
     {'Quantized'     }          19.433                       64                                   16                            62.31                     50.11                   32.103       

Import a dlquantizer Object into the Deep Network Quantizer App

This example shows you how to import a dlquantizer object from the base workspace into the Deep Network Quantizer app. This allows you to begin quantization of a deep neural network using the command line or the app, and resume your work later in the app.

Open the Deep Network Quantizer app.

deepNetworkQuantizer

In the app, click New and select Import dlquantizer object.

In the dialog, select the dlquantizer object to import from the base workspace. For this example, use quantObj that you create in the above example Quantize a Neural Network for GPU Target.

The app imports any data contained in the dlquantizer object that was collected at the command line. This data can include the network to quantize, calibration data, validation data, and calibration statistics.

The app displays a table containing the calibration data contained in the imported dlquantizer object, quantObj. To the right of the table, the app displays histograms of the dynamic ranges of the parameters. The gray regions of the histograms indicate data that cannot be represented by the quantized representation. For more information on how to interpret these histograms, see Quantization of Deep Neural Networks.

Emulate Target Agnostic Quantized Network

This example shows how to create a target agnostic, simulatable quantized deep neural network in MATLAB.

Target agnostic quantization allows you to see the effect quantization has on your neural network without target hardware or target-specific quantization schemes. Creating a target agnostic quantized network is useful if you:

• Do not have access to your target hardware.

• Want to preview whether or not your network is suitable for quantization.

• Want to find layers that are sensitive to quantization.

Quantized networks emulate quantized behavior for quantization-compatible layers. Network architecture like layers and connections are the same as the original network, but inference behavior uses limited precision types. Once you have quantized your network, you can use the quantizationDetails function to retrieve details on what was quantized.

Load the pretrained network. net is a SqueezeNet network that has been retrained using transfer learning to classify images in the MerchData data set.

load squeezenetmerch
net

net = 
  DAGNetwork with properties:

         Layers: [68×1 nnet.cnn.layer.Layer]
    Connections: [75×2 table]
     InputNames: {'data'}
    OutputNames: {'new_classoutput'}

You can use the quantizationDetails function to see that the network is not quantized.

qDetailsOriginal = quantizationDetails(net)

qDetailsOriginal = struct with fields:
            IsQuantized: 0
          TargetLibrary: ""
    QuantizedLayerNames: [0×0 string]
    QuantizedLearnables: [0×3 table]

Unzip and load the MerchData images as an image datastore.

unzip('MerchData.zip')
imds = imageDatastore('MerchData', ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');

Define calibration and validation data to use for quantization. The output size of the images are changed for both calibration and validation data according to network requirements.

[calData,valData] = splitEachLabel(imds,0.7,'randomized');
augCalData = augmentedImageDatastore([227 227],calData);
augValData = augmentedImageDatastore([227 227],valData);

Create dlquantizer object and specify the network to quantize. Set the execution environment to MATLAB. How the network is quantized depends on the execution environment. The MATLAB execution environment is agnostic to the target hardware and allows you to prototype quantized behavior.

quantObj = dlquantizer(net,'ExecutionEnvironment','MATLAB');

Use the calibrate function to exercise the network with sample inputs and collect range information. The calibrate function exercises the network and collects the dynamic ranges of the weights and biases in the convolution and fully connected layers of the network and the dynamic ranges of the activations in all layers of the network. The function returns a table. Each row of the table contains range information for a learnable parameter of the optimized network.

calResults = calibrate(quantObj,augCalData);

Attempt to calibrate with host GPU errored with the message: 
Unable to find a supported GPU device. For more information on GPU support, see GPU Support by Release. 
Reverting to use host CPU. 

Use the quantize method to quantize the network object and return a simulatable quantized network.

qNet = quantize(quantObj)  

qNet = 
  Quantized DAGNetwork with properties:

         Layers: [68×1 nnet.cnn.layer.Layer]
    Connections: [75×2 table]
     InputNames: {'data'}
    OutputNames: {'new_classoutput'}

  Use the quantizationDetails function to extract quantization details.

You can use the quantizationDetails function to see that the network is now quantized.

qDetailsQuantized = quantizationDetails(qNet)

qDetailsQuantized = struct with fields:
            IsQuantized: 1
          TargetLibrary: "none"
    QuantizedLayerNames: [26×1 string]
    QuantizedLearnables: [52×3 table]

Make predictions using the original, single-precision floating-point network, and the quantized INT8 network.

predOriginal = classify(net,augValData);       % Predictions for the non-quantized network
predQuantized = classify(qNet,augValData);     % Predictions for the quantized network 

Compute the relative accuracy of the quantized network as compared to the original network.

ccrQuantized = mean(predQuantized == valData.Labels)*100

ccrQuantized = 100

ccrOriginal = mean(predOriginal == valData.Labels)*100

ccrOriginal = 100

For this validation data set, the quantized network gives the same predictions as the floating-point network.

Version History

Introduced in R2020a

R2023b: Simulation property of dlquantizer removed

The Simulation property has been removed from dlquantizer. When you set the ExecutionEnvironment property to "FPGA", the software validates the quantized network through emulation by default. For hardware validation, set the Target property of dlquantizationOptions to dlhdl.Target.

R2023a: dlnetwork support

dlquantizer now supports dlnetwork objects for quantization using the quantize function.

R2023a: Quantize yolov3ObjectDetector and yolov4ObjectDetector using dlquantizer

yolov3ObjectDetector (Computer Vision Toolbox) and yolov4ObjectDetector (Computer Vision Toolbox) objects can now be quantized using dlquantizer.

R2022b: dlnetwork support

dlquantizer now supports dlnetwork objects for quantization using the calibrate and validate functions.

R2022a: Validate the performance of quantized network for CPU target

You can now use the dlquantizer object and the validate function to quantize a network and generate code for CPU targets.

R2022a: Quantize neural networks without a specific target

Specify MATLAB as the ExecutionEnvironment to quantize your neural networks without generating code or committing to a specific target for code deployment. This can be useful if you:

• Do not have access to your target hardware.

• Want to inspect your quantized network without generating code.

Your quantized network implements int8 data instead of single data. It keeps the same layers and connections as the original network, and it has the same inference behavior as it would when running on hardware.

Once you have quantized your network, you can use the quantizationDetails function to inspect your quantized network. Additionally, you also have the option to deploy the code to a GPU target.