Answered
semanticseg - out of memory on device
I understand that there's a bug in this demo script, and the default MiniBatchSize is too large and should be reduced to 4: ...

8 years ago | 0

| accepted

Answered
help error CNN The output of layer 6 is incompatible with the input expected by layer 7.
I ran the <https://uk.mathworks.com/help/nnet/ref/analyzenetwork.html Network Analyzer> on your network: analyzeNetwork(lay...

8 years ago | 0

Answered
Solving sparse matrix on GPU and memory problems (even with free memory available)
Hi Paulo. I must admit, I'm not extremely familiar with the behaviour of the matrix factorization we use to implement the sparse...

8 years ago | 0

Answered
GPU Array lazy evaluation?
MATLAB doesn't take 200MB on device selection, the CUDA driver does. MATLAB pools up to a quarter of the GPU memory by defaul...

8 years ago | 1

Answered
Is Matlab/ Simulink supporting Nvidia external GPU Systems/Thunderbold 3 for Machine Learning and Deep Learning projects including embedded systems like Jetson TX2?
Yes, it's supported. This isn't really a question of MATLAB's system requirements; if the CUDA driver can interact with your car...

8 years ago | 2

| accepted

Answered
GPU overload - specifying gpus using parpool
That code is selecting device 2 on every worker. You want something like parpool('local', 3); spmd gpuDevice(labi...

8 years ago | 1

| accepted

Answered
How to start operations in parrallel once the data has become avaliable
Try <https://www.mathworks.com/help/distcomp/parfeval.html parfeval> in conjunction with <https://www.mathworks.com/help/distcom...

8 years ago | 1

| accepted

Answered
why iam having error "The following error occurred converting from gpuArray to single: Conversion to single from gpuArray is not possible"
This can happen sometimes when you are running low on GPU memory due to a bug. You will be receiving the "GPULowOnMemory" warnin...

8 years ago | 1

Answered
call mexcuda file returned gpuarray cost abnormal time
The time is all in your mex function. It looks as though the time is spent in line 11 because your kernel is running asynchronou...

8 years ago | 1

| accepted

Answered
Connection problems with cloud center to AWS p2.8 GPU
P2 is limited and under heavy use. You should try a range of different regions, see what's available. You could also try P3, G2 ...

8 years ago | 0

Answered
Manipulating matrices/vectors in GPU array operation
It's hard to know how to get started answering your question. Using the GPU requires vectorized code, so you need to translate l...

8 years ago | 0

Answered
CUDA Unexpected Error for nndata2gpu
Looks like you found a bug, many thanks. We will investigate. Meanwhile, best guess for now, this is caused by using more data t...

8 years ago | 0

| accepted

Answered
interp1() broken on GPUs in 2017a?
This is a bug, thanks for reporting it to us. To work around, make sure the data X and query Xq are the same datatype.

8 years ago | 0

| accepted

Answered
error while trying to train network using multiple cpus
The error means that the |vdsrImagePatchDatastore| does not support parallel training. This is a custom datastore created for th...

8 years ago | 0

| accepted

Answered
What does the error message "Requested code-table for unsupported compute-capability: 2" mean?
It means you've hit a <https://uk.mathworks.com/support/bugreports/search_results?utf8=✓&search_executed=1&keyword=Fermi&release...

8 years ago | 0

| accepted

Answered
If gpu computing or convert to mex file help to accelerate the code
It doesn't look as though the problem is the speed of execution of those functions, so much as the number of times they are call...

8 years ago | 0

| accepted

Answered
How is it possible to use a validation set with a LSTM?
It's ugly, but if you use Checkpoints, then you can use an OutputFcn to (once per epoch) load the network from a checkpoint and ...

8 years ago | 1

| accepted

Answered
mex/mexcuda for a target GPU
The supported (if not exactly documented) way of doing this is to define the variable |NVCCFLAGS| in your call to |mexcuda|. ...

8 years ago | 1

| accepted

Answered
Parallel running coroutines in Matlab
This is one of the major purposes of the Parallel Computing Toolbox. Start by taking a look at <https://uk.mathworks.com/help/di...

8 years ago | 0

| accepted

Answered
Is there a plan to have gpu support for rangesearch?
The best I can say is: I have reported your request to the Stats Team. So I hope you'll have it in a future release!

8 years ago | 0

Answered
Processing on GPU in Simulink
|gpuArray| is not yet supported by GPU Coder. To see how you can use GPU Coder with Simulink, take a look at <https://uk.mathwor...

8 years ago | 0

| accepted

Answered
GPU speed up, matrix multiplication
|tic| and |toc| are not giving the correct timings for your first set of iterations, because your kernels are launching asynchro...

8 years ago | 0

Answered
slow training on single gpu
Upgrade MATLAB with each new release, we are making big performance improvements all the time.

8 years ago | 0

Answered
GPU programming on Matlab (Mac/Window): What are my options?
Others may have a longer answer to your question, but certainly it is possible to get an NVIDIA GPU for a desktop PC with a PCI ...

8 years ago | 2

| accepted

Answered
Use CUdA and CudNN with Matlab
Check <https://uk.mathworks.com/help/gpucoder/gs/setting-up-the-toolchain.html some of these environment variables> to make sure...

8 years ago | 0

Answered
Deep Learning - Data no longer exists on the device
This would happen if you are resetting the device or switching device using |reset| or |gpuDevice|. Perhaps you have a custom la...

8 years ago | 2

Answered
Code Vectorization in custom layer
Adotf = cellfun(@(aa,ff)ff.'*aa, A, f, 'UniformOutput', false); Adotf = cat(1, Adotf{:}); Z = X(k).*Adotf(jj,:); j = ...

8 years ago | 0

| accepted

Answered
Dear friends, when I use the gather command of MATLAB 2016a， sometimes this command takes a very large amount of time. That's why？ thanks a lot.
|gather| itself cannot possibly take 5 seconds - you can prove this by running |gputimeit(@()gather(b))| rather than using |tic|...

8 years ago | 1

Answered
I need to do this without for loop(For Gpu computing)
You can't do this, but I will take it as a further request for |pagefun| to support |qr| which will be coming in a future versio...

8 years ago | 1

| accepted

Answered
MATLAB Requirements for Neural Network on GPU
Yes.

8 years ago | 0

Load more