pre-trained model lacking fields lead to error

[BigBugX]

Hi all,

I've met a similar problem, and do not know what to do with it. The error info is as follows:

train: epoch 01: 1/ 11:Error using dagnn.DagNN/eval (line 80)
No variable of name 'input' could be found in the DAG.

Error in cnn_train_dag>process_epoch (line 213)
net.eval(inputs, opts.derOutputs) ;

Error in cnn_train_dag (line 83)
[stats.train(epoch),prof] = process_epoch(net, state, opts, 'train') ;

Error in cnn_four (line 61)
[net, info] = trainFn(net, imdb, getBatchFn(opts, net.meta), ...

I use the pre-trained model "imagenet-matconvnet-alex.mat", and found that when calling "eval", the net's lacking of some fields results the error. "net"(or "obj" in def. of "eval") lacking fields like "varNames", "numPendingVarRefs", ... . Does any body know how to fix it?

Thanks and Regards