How to solve these problems?

Issue #24 closed
Ding Zheyuan
created an issue

Dear Tsungyu:    Thanks for your sharing at first! I have met some problems when I want to run these files such as bird_demo , model_train and run_experiments_bcnn_train. Each problems likes below.

       1 bird_demo:
       0.10s to load imdb.

Warning: While loading an object of class 'dagnn.DagNN': Undefined function or variable 'BilinearClPooling'.

In bird_demo (line 30) Reference to non-existent field 'value'.

Error in dagnn.DagNN/move (line 22) obj.params(i).value = gather(obj.params(i).value) ; Error in net_move_to_device (line 6) net.move(device)

Error in bird_demo (line 34) net = net_move_to_device(net, 'cpu'); 2. model_train error as this:

Index exceeds matrix dimensions.

Error in model_train>traintest (line 133) [w{c},b{c}] = vl_svmtrain(psi(:,train & y ~= 0), y(train & y ~= 0), 1/(n* C), ...

     3. run_experiments_bcnn_train errors:

train: epoch 01: 1/ 24:Error using load Unable to read file 'data/checkgpu/cub-seed-01/nonftbcnn/bcnn_nonft_07031'. No such file or directory.

Error in initializeNetworkSharedWeights>getBatch_bcnn_fromdisk (line 201) load(fullfile(imdb.imageDir,{batch(i)}));

Error in cnn_train>processEpoch (line 291) [im, labels] = params.getBatch(, batch) ;

Error in cnn_train (line 132) [net, state] = processEpoch(net, state, params, 'train') ;

Error in initializeNetworkSharedWeights (line 161) [netc, info] = cnn_train(netc, bcnndb, @getBatch_bcnn_fromdisk, opts.inittrain, ...

Error in imdb_bcnn_train_dag (line 65) net = initNetFn(imdb, encoderOpts, opts);

Error in run_experiments_bcnn_train (line 79) imdb_bcnn_train_dag(imdb, opts);

       I'm sorry to trouble you, but I really have no idea where the problem is. I'm looking forward to hearing from you.
      Best wishes for you!

Comments (3)

  1. tsungyu repo owner

    Please check if you have the submodule set up properly. There should be several files including 'BilinearClPooling.m' in the folder "bcnn-package".

    For the error "unable to read data/checkgpu/cub-seed-01/nonftbcnn/bcnn_nonft_07031", please see Issue 21

    You should not run the model_train.m directly.

  2. Ding Zheyuan reporter

    Thank you very much.Now I can train it.I should run the run_experiments_train.m at first and then use the model after training to run the run_experiments_bcnn_train file.

  3. Log in to comment