After running " python cdl.py" I have this problem

[daicoolb]

            Hi,After running " python cdl.py" I have this problem：
            Traceback (most recent call last):
            File "cdl.py", line 61, in 
            lr_scheduler=mx.misc.FactorScheduler(20000,0.1))
            File "/home/daicoolb/MXNet-for-CDL/autoencoder.py", line 175, in finetune
            0, n_iter, {}, False)
            File "/home/daicoolb/MXNet-for-CDL/solver.py", line 84, in solve
            update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_a rrays) if nd}
            File "/home/daicoolb/MXNet-for-CDL/solver.py", line 84, in 
            update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_a rrays) if nd}
            File "/usr/local/lib/python2.7/dist-packages/mxnet/ndarray.py", line 238, in _ bool_
            raise ValueError("The truth value of an NDArray with more than one element i s ambiguous.")
            ValueError: The truth value of an NDArray with more than one element is ambiguou s.

[nimit2894]

I am facing the same issue, and I would really appreciate the creator of this repo to respond to this, as it is quite hard to find out the source of this error from the huge ndarray.py file that has been created in mxnet

[js05212]

Hi, sorry for the late reply. I think this is because you are using a newer version of MXNet. There might be some compatibility issue in here. Have you try using earlier version of MXNet?

[xanhho]

Hi, I also have the same error when I running cdl.py INFO:root:Fine tuning... Traceback (most recent call last): File "", line 4, in File "autoencoder.py", line 175, in finetune return U, V, theta, BCD_loss File "solver.py", line 84, in solve update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_arrays) if nd} File "solver.py", line 84, in update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_arrays) if nd} File "/home/xanhho/.local/lib/python2.7/site-packages/mxnet/ndarray.py", line 293, in bool raise ValueError("The truth value of an NDArray is ambiguous. " \ ValueError: The truth value of an NDArray is ambiguous. Please convert to number with asscalar() first.

Can you guide me how to fix it?

[js05212]

Hi, might I ask what version of MXNet you are using?

[xanhho]

Thank you for your reply! I had fixed this error by install another version of MXNet. I have some question related to your code. Could you please answer for me? In the paper had mentioned 5 hyperparameters lambda_w, lambda_n, lambda_u, lambda_s, lambda_v. But I see you just mentioned lambda_u and lambda_v in your code? How about lambda_w, lambda_n, lambda_s? Is there any related between lv, lambda_v_rt (in your code) and lambda_w, lambda_n, lambda_s (in the paper)?

Thank you very much!

[js05212]

Hi. lambda_w is equivalent to the weight decay coefficient in NN if we use MAP inference, and it is fixed to 1e-4 (since the performance is not sensitive to it). lambda_s can be taken to positive infinity as mentioned in the paper. Essentially what is important is the ratio lambda_n/lambda_v and lambda_u.

Hope this helps!

[xanhho]

Hi, thank you for your reply!

Is this code can reproduce the result as mentioned in the paper? Or it just a simple version for CDL?

[js05212]

Hi, it is a simple version for CDL. For the full version, please go to wanghao.in/code/cdl-release.rar.You are more than welcome to fork this repo to make it the full version though. Feel free to ask if you need help.

[xanhho]

Hi, thank you very much for your support!

[m2rik]

hi, can i know which version of MXNet solved the problem ?

[xanhho]

Hi, I am using MXNet 0.11.0.

[m2rik]

i'm using the same version still getting the same error ,any solution ?

[xanhho]

Can you show me the error? Have you tried debug by python command?

[m2rik]

[xanhho]

You can fix this error by change the code in line 84 by this line:

update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_arrays) if nd}

    update_dict = {name: nd for name, nd in zip(sym.list_arguments(), exe.grad_arrays) if nd is not None}

[m2rik]

Okay,all the errors were just the Python version issue .Now stuck on the show recommendations file with the error attribute error object has no attribute 'A1' .can anyone solve this

[xanhho]

Change by these following lines, it will work. // s_test = set(np.where(R_test[user_id,:]>0)[1].A1) // l_train = np.where(R_train[user_id,:]>0)[1].A1.tolist() s_test = set(np.ravel(np.where(R_test[user_id, :] > 0)[1])) l_train = np.ravel(np.where(R_train[user_id, :] > 0)[1]).tolist()

[m2rik]

It worked, can you tell me how it fixes the A1 attribute of the array?

[js05212]

Thanks, @xanhxanh94, for your help!

[xanhho]

Hi, @js05212, You're welcome! Can you please answer for me the questions in js05212/CDL?

Thank you very much for your support!

[ShuyaZhang]

Hi, @xanhxanh94 , I've corrected several mistakes you mentioned before, now I'm using earlier version of mxnet. However, I still have following error when running cdl.py, could you solve it?

[MaurizioFD]

Hi, @xanhxanh94 , I've corrected several mistakes you mentioned before, now I'm using earlier version of mxnet. However, I still have following error when running cdl.py, could you solve it?

The error is due to the fact that in python 3 the .values() function returns a view.

You can solve it by creating a list with the dict_values object. The code in solver.py in lines 131-138 becomes:

theta = list(model.extract_feature(sym[0], args, auxs,
    data_iter, X.shape[0], xpu).values())[0]
# update U, V and get BCD loss
U, V, BCD_loss = BCD_one(R, U, V, theta,
    lambda_u, lambda_v, dir_save, True)
# get recon' loss
Y = list(model.extract_feature(sym[1], args, auxs,
    data_iter, X.shape[0], xpu).values())[0]