dlquantizationOptions

Options for quantizing a trained deep neural network

Description

The dlquantizationOptions object provides options for quantizing a trained deep neural network to scaled 8-bit integer data types.

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

Creation

Syntax

quantOpts = dlquantizationOptions

quantOpts = dlquantizationOptions(Name,Value)

Description

quantOpts = dlquantizationOptions creates a dlquantizationOptions object with default property values.

quantOpts = dlquantizationOptions(Name,Value) creates a dlquantizationOptions object with additional properties specified by one or more name-value pair arguments.

example

Properties

expand all

`MetricFcn` — Metric function to use for validation of quantized network
cell array of function handles

Metric function to use for validation of quantized network, specified as a cell array of one or more function handles.

Example: options = dlquantizationOptions('MetricFcn',{@(x)hComputeModelAccuracy(x,net,groundTruth)});

Data Types: cell

Execution Environment Options

`Bitstream` — Name of FPGA bitstream
`'zcu102_int8'` | `'zc706_int8'` | `'arria10soc_int8'`

This property is valid only when the 'ExecutionEnvironment property of the dlquantizer object is set to 'FPGA'.

Name of the FPGA bitstream, specified as one of these values or as the path to a custom bitstream:

Bitstream	Target Board
`'zcu102_int8'`	Xilinx^® Zynq^® UltraScale™ ZCU102
`'zc706_int8'`	Xilinx Zynq-7000 ZC706
`'arria10soc_int8'`	Intel^® Arria^® 10 SoC development kit

You can specify a custom bitstream for any supported target board or for a custom target. To use a custom bitstream, first specify a dlhdl.Target object, then specify the path to a valid bitstream file ending in .sof or .bit depending on your target board. For more information about generating custom bitstreams, see Generate Custom Bitstream

Example: quantOpts = dlquantizationOptions('Bitstream','zcu102_int8')

Example: hTarget = dlhdl.Target('Intel','Interface','JTAG'); quantOpts = dlquantizationOptions('Target',hTarget,'Bitstream','C:\yourFolder\customBitstream_int8.bit')

`Target` — Target for quantized network
`'host'`(default) | `'gpu'` | `dlhdl.Target` object | `raspi` object

Target for quantized network, specified as one of the following:

Target	Execution Environment for Quantized Network	Example
Quantized network in MATLAB specified with `'host'`	Set `Target` property as `'host'` when `'ExecutionEnvironment'` property of the `dlquantizer` object is set to `'GPU'`,`'FPGA'`, or `'MATLAB'`	`quantOpts = dlquantizationOptions('Target','host')`
Target GPU device specified with `'gpu'`	Set `Target` property as `'gpu'` only when `'ExecutionEnvironment'` property of the `dlquantizer` object is set to `'GPU'`	`quantOpts = dlquantizationOptions('Target','gpu')`
Target CPU board specified as a `raspi` object	Set `Target` property as a `raspi` object only when `'ExecutionEnvironment'` property of the `dlquantizer` object is set to `'CPU'`	`r = raspi('hostname','User Name','Password'); quantOpts = dlquantizationOptions('Target',r)`
Target FPGA board vendor name and interface, specified as a `dlhdl.Target` object	Set `Target` property as a `dlhdl.Target` object only when `'ExecutionEnvironment'` property of the `dlquantizer` object is set to `'FPGA'`	`hTarget = dlhdl.Target('Intel','Interface','JTAG'); quantOpts = dlquantizationOptions('Target',hTarget)`

Examples

collapse all

Quantize a Neural Network for GPU Target

This example uses:

Open Live Script

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. 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 squeezedlnetmerch
net

net = 
  dlnetwork with properties:

         Layers: [67×1 nnet.cnn.layer.Layer]
    Connections: [74×2 table]
     Learnables: [52×3 table]
          State: [0×3 table]
     InputNames: {'data'}
    OutputNames: {'prob'}
    Initialized: 1

  View summary with summary.

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');
classes = categories(imds.Labels);
[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');
quantOpts.MetricFcn = {@(x)hAccuracy(x,net,aug_valData,classes)}

quantOpts = 
  dlquantizationOptions with properties:

   Validation Metric Info
    MetricFcn: {[@(x)hAccuracy(x,net,aug_valData,classes)]}

   Validation Environment Info
       Target: 'gpu'
    Bitstream: ''

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=120×5 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 Network for FPGA Deployment

This example uses:

Open Live Script

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 Compression Library and Deep Learning HDL Toolbox to deploy the int8 network to a target FPGA board.

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 object and specify the network to quantize. Specify the execution environment as FPGA.

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

Use the calibrate 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

Quantize a Neural Network for CPU Target

This example uses:

Open Live Script

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 squeezedlnetmerch
net

net = 
  dlnetwork with properties:

         Layers: [67×1 nnet.cnn.layer.Layer]
    Connections: [74×2 table]
     Learnables: [52×3 table]
          State: [0×3 table]
     InputNames: {'data'}
    OutputNames: {'prob'}
    Initialized: 1

  View summary with summary.

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

unzip('MerchData.zip');
imds = imageDatastore('MerchData', ...
    'IncludeSubfolders',true, ...
    'LabelSource','foldernames');
classes = categories(imds.Labels);
[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 dlnetwork]
    ExecutionEnvironment: 'CPU'

Calibrate the network.

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

calResults=120×5 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-8','pi','matlab')

r = 
  raspi with properties:

         DeviceAddress: 'gpucoder-raspberrypi-8'      
                  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.MetricFcn = {@(x)hAccuracy(x,net,aug_valData,classes)}

opts = 
  dlquantizationOptions with properties:

   Validation Metric Info
    MetricFcn: {[@(x)hAccuracy(x,net,aug_valData,classes)]}

   Validation Environment Info
       Target: [1×1 raspi]
    Bitstream: ''

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'}           1      
     {'Quantized'     }           1

Version History

Introduced in R2020a

expand all

R2023a: Specify Raspberry Pi as quantization target

You can now specify a raspi object as the target for quantization using the Target property when dlquantizer Execution Environment is set to CPU.

dlquantizationOptions

Description

Creation

Syntax

Description

Properties

`MetricFcn` — Metric function to use for validation of quantized network
cell array of function handles

Execution Environment Options

`Bitstream` — Name of FPGA bitstream
`'zcu102_int8'` | `'zc706_int8'` | `'arria10soc_int8'`

`Target` — Target for quantized network
`'host'`(default) | `'gpu'` | `dlhdl.Target` object | `raspi` object

Examples

Quantize a Neural Network for GPU Target

Quantize Network for FPGA Deployment

Quantize a Neural Network for CPU Target

Version History

R2023a: Specify Raspberry Pi as quantization target

See Also

Apps

Functions

Topics

dlquantizationOptions

Description

Creation

Syntax

Description

Properties

MetricFcn — Metric function to use for validation of quantized network cell array of function handles

Execution Environment Options

Bitstream — Name of FPGA bitstream 'zcu102_int8' | 'zc706_int8' | 'arria10soc_int8'

Target — Target for quantized network 'host'(default) | 'gpu' | dlhdl.Target object | raspi object

Examples

Quantize a Neural Network for GPU Target

Quantize Network for FPGA Deployment

Quantize a Neural Network for CPU Target

Version History

R2023a: Specify Raspberry Pi as quantization target

See Also

Apps

Functions

Topics

`MetricFcn` — Metric function to use for validation of quantized network
cell array of function handles

`Bitstream` — Name of FPGA bitstream
`'zcu102_int8'` | `'zc706_int8'` | `'arria10soc_int8'`

`Target` — Target for quantized network
`'host'`(default) | `'gpu'` | `dlhdl.Target` object | `raspi` object