deep network quantizer does not work

Hi,

I have istalled the add-on and the required toolbox. However, when I hit "New" in Deep Network Quantizer, I get the error message "Deep Network Quantizer requires a GPU execution environment. How to fix it?

I'm using Matlab 2020a. This is my GPU settings:

>> gpuDevice

ans =

CUDADevice with properties:

Name: 'GeForce GTX 1070'

Index: 1

ComputeCapability: '6.1'

SupportsDouble: 1

DriverVersion: 11

ToolkitVersion: 10.1000

MaxThreadsPerBlock: 1024

MaxShmemPerBlock: 49152

MaxThreadBlockSize: [1024 1024 64]

MaxGridSize: [2.1475e+09 65535 65535]

SIMDWidth: 32

TotalMemory: 8.5899e+09

AvailableMemory: 7.0146e+09

MultiprocessorCount: 16

ClockRateKHz: 1645000

ComputeMode: 'Default'

GPUOverlapsTransfers: 1

KernelExecutionTimeout: 1

CanMapHostMemory: 1

DeviceSupported: 1

DeviceSelected: 1

Answer 1

[Raymond Norris]

Hi Kaijiun,

Looks like you have the right ComputeCapability (6.1). Just a quick check: in addition to the Deep Learning Toolbox Model Quantization Library support package, you also have the GPU Coder Interface for Deep Learning Libraries support package installed as well, is that right?

If so, I would suggest reaching out to Technical Support ([e-mail]) to have them troubleshoot this.

Raymond

[Kaijiun Yang]

Hi, Mr. Norris,

Thanks for your help.

Yes, I have both libraries installed. However, the problem is as I've stated.

I'll write the mail for further help.

[Ali Al-Saegh]

Dear Raymond,

Please help me. I posted a question here https://www.mathworks.com/matlabcentral/answers/734902-size-of-a-quantized-dnn?s_tid=srchtitle .

Answer 2

[ikeuwanuakwa]

I have a different problem, this line of code is given error

quantOpts = dlquantizationOptions('MetricFcn', ...

{@(x)hComputeModelAccuracy(x, net, aug_valData)});

calResults = calibrate(quantObj, aug_calData)

valResults = validate(quantObj, aug_valData, quantOpts)

valResults.MetricResults.Result

Error

Function definitions in a script must appear at the end of the file.

Move all statements after the "hComputeModelAccuracy" function definition to before the first local function

definition.

[Raymond Norris]

See https://www.mathworks.com/matlabcentral/answers/392662-function-definitions-must-appear-at-end-of-file