peiyunh commented 8 years ago

CuDNN v5.1 speeds up 3x3 convolution substantially.

Now I'm using MATLAB R2015b with MatConvNet beta19. I was able to compile it but saw errors at runtime:

Error using vl_nnconv
forward: cuDNN error [cudnn:
  #@"~/matconvnet/matlab/src/bits/impl/nnconv_cudnn.cu":118
(CUDNN_STATUS_BAD_PARAM)]

Error in dagnn.Conv/forward (line 11)
      outputs{1} = vl_nnconv(...

Error in dagnn.Layer/forwardAdvanced (line 85)
      outputs = obj.forward(inputs, {net.params(par).value}) ;

Error in dagnn.DagNN/eval (line 88)
  obj.layers(l).block.forwardAdvanced(obj.layers(l)) ;

Error in main>evaluate (line 266)
    net.eval(inputs(1:2));

I'm going to test with beta20 to see if it fixes the issue.

aaatingtingli commented 7 years ago

hello, I met the same problem yet. Have you solved it?

daa233 commented 6 years ago

I have met the same problem for a few days, have any one solved this?

vlfeat / matconvnet

Use CuDNN v5.1 #611

CuDNN v5.1 speeds up 3x3 convolution substantially.