pscedu / matconvnet

MatConvNet: CNNs for MATLAB

MatConvNet: CNNs for MATLAB

MatConvNet is a MATLAB toolbox implementing Convolutional Neural Networks (CNNs) for computer vision applications. It is simple, efficient, and can run and learn state-of-the-art CNNs. Several example CNNs are included to classify and encode images. Please visit the homepage to know more.

In case of compilation issues, please read first the Installation and FAQ section before creating an GitHub issue. For general inquiries regarding network design and training related questions, please use the Discussion forum.

Compiling on Bridges-2

Last Update: April 22, 2022

1. Start an interactive session. Note: if you want to have GPU and test it, you need to both load CUDA and request a GPU.

interact -gpu -n 5 --mem=11Gb

2. Load the appropriate modules

module load matlab/R2021a
module load cuda/11.1.1

3. Clone repo

git clone [e-mail]:vlfeat/matconvnet.git

4. Edit Makefile file Inside the matconvnet folder you will see a file named Makefile. Change line 27 from

CUDAROOT ?= /usr/local/cuda

to

CUDAROOT ?= /jet/packages/spack/opt/spack/linux-centos8-zen/gcc-8.3.1/cuda-11.1.1-a6ajxenobex5bvpejykhtnfut4arfpwh

6. Change directories to the folder where the file Makefile lives and run the command

make

it might take several minutes to compile. Be patient.

7. Start Matlab to confirm it compiled and works properly. Following the instructions in this package

matlab -nodesktop -nosplash

MATLAB is selecting SOFTWARE OPENGL rendering.

                                                                < M A T L A B (R) >
                                                      Copyright 1984-2021 The MathWorks, Inc.
                                                 R2021a Update 1 (9.10.0.1649659) 64-bit (glnxa64)
                                                                   April 13, 2021

 
To get started, type doc.
For product information, visit www.mathworks.com.
 
>> run matlab/vl_setupnn.m
>> vl_testnn

 Running nnbilinearsampler
  Setting up nnbilinearsampler[dataType=single,device=cpu]
  Done setting up nnbilinearsampler[dataType=single,device=cpu] in 0.000941 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value1)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value1) in 0.67398 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value2)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value1,multiple_grids=value2) in 0.076365 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value2,multiple_grids=value1)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value2,multiple_grids=value1) in 0.077341 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value2,multiple_grids=value2)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value1,oh=value1,ow=value2,multiple_grids=value2) in 0.061363 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value1,multiple_grids=value1)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value1,multiple_grids=value1) in 0.093338 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value1,multiple_grids=value2)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value1,multiple_grids=value2) in 0.036754 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value2,multiple_grids=value1)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value2,multiple_grids=value1) in 0.034694 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value2,multiple_grids=value2)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value2,oh=value1,ow=value2,multiple_grids=value2) in 0.039541 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value3,oh=value1,ow=value1,multiple_grids=value1)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value3,oh=value1,ow=value1,multiple_grids=value1) in 0.030887 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value3,oh=value1,ow=value1,multiple_grids=value2)
   Done nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value3,oh=value1,ow=value1,multiple_grids=value2) in 0.036112 seconds
   Running nnbilinearsampler[dataType=single,device=cpu]/check_der_x(ih=value1,iw=value3,oh=value1,ow=value2,multiple_grids=value1)
.... (suppressing most of the output)

result = 

  1x1793 TestResult array with properties:

    Name
    Passed
    Failed
    Incomplete
    Duration
    Details

Totals:
   1793 Passed, 0 Failed, 0 Incomplete.
   124.5039 seconds testing time.

The tests will take several minutes as well. Be patient.