Error in training phase

[cndylan]

I try to train the network but meeting an error.

train: epoch 01:   1/5985:Error using vl_nnbnorm
The MOMENTS size does not match the DATA depth.

Error in dagnn.BatchNorm/backward (line 29)
        vl_nnbnorm(inputs{1}, params{1}, params{2}, derOutputs{1}, ...

Error in dagnn.Layer/backwardAdvanced (line 120)
      [derInputs, derParams] = obj.backward ...

Error in dagnn.DagNN/eval (line 117)
  obj.layers(l).block.backwardAdvanced(obj.layers(l)) ;

Error in cnn_train_dag>processEpoch (line 253)
      net.eval(inputs, params.derOutputs, 'holdOn', s < params.numSubBatches) ;

Error in cnn_train_dag (line 105)
    [net, state] = processEpoch(net, state, params, 'train') ;

Error in experiment (line 102)
    [net, stats] = cnn_train_dag(net, imdb, batch_fn, opts.train);

I use matconvnet-1.0-beta25.
can any help provided ?
Thanks

[bertinetto]

Hi, unfortunately we are not able to support newer versions of Matconvnet. Does this error arise from the code as it is or did you modify something?

[xukai8]

I met the same problem. Have you solved the problem?

[nanzxl]

I simply remove the line 33 in BatchNorm.m
% obj.moments=[ ]; Hope this helps.

[moorejee]

Is this problem caused by the different versions? I want to train the model using the last version, I am afraid of the same error.

[NeverMoreLCH]

I met the same problem. Have you solved the problem?

[NeverMoreLCH]

Oh, I might solve the problem so that the training process can start. But I don't know if some other questions will occur.

I change the code in BatchNorm.m. At line 32, remove the parameter 'moments'. And remove the line 33.

[ZZXin]

Oh, I might solve the problem so that the training process can start. But I don't know if some other questions will occur.

I change the code in BatchNorm.m. At line 32, remove the parameter 'moments'. And remove the line 33.

Solved my problem, hope not affect the final training results.