kurowasan / gran-dag Goto Github PK

After installed Singularity, I changed the start_example.sh setting, and put DAG1.npy ... files to a path. But here's what I get:

$ sudo sh start_example.sh
No GPU automatically detected. Setting SETTINGS.GPU to 0, and SETTINGS.NJOBS to cpu_count.
sklearn.externals.joblib is deprecated in 0.21 and will be removed in 0.23. Please import this functionality directly from joblib, which can be installed with: pip install joblib. If this warning is raised when loading pickled models, you may need to re-serialize those models with scikit-learn 0.21+.
Traceback (most recent call last):
  File "main.py", line 133, in <module>
    main(parser.parse_args())
  File "/code/gran_dag/main.py", line 87, in main
    normalize=opt.normalize_data, random_seed=opt.random_seed)
  File "/code/gran_dag/data.py", line 43, in __init__
    adjacency = np.load(os.path.join(file_path, "DAG{}.npy".format(i_dataset)))
  File "/usr/local/lib/python3.7/dist-packages/numpy/lib/npyio.py", line 384, in load
    fid = open(file, "rb")
FileNotFoundError: [Errno 2] No such file or directory: '/home/xxx/GraN-DAG/dataset/DAG1.npy'

$ ls -l /home/xxx/GraN-DAG/dataset/DAG1.npy
-rw-r--r-- 1 xxx xxx 80128 May  4 08:49 /home/xxx/GraN-DAG/dataset/DAG1.npy

It seems to find data in the docker, cause I dont have python3.7 in my ubuntu.
How to deal with that?

Hi,
I really like your paper and was wondering if one could, similarly to DAG-GNN use discrete variables as well? That is, instead of outputting the mean and standard deviation for each continuous variable, could one output a discrete distribution like the factored categorical in DAG-GNN? Would that be possible in theory and would one need to change the loss function much?

Hi, I want to run GraN-DAG with my dataset. However, as soon as I tried to open the link to the container, google drive returned me "Page not found".

I loved reading your paper!

Hi there,

Sorry for bothering again, I am implementing the Pruning module solely (simply pass the data matrix and estimated DAG into the def cam_pruning_). Logically, this should work right? But it returns all Zeros matrix which mean this function prunes all the edges. (All the R dependencies are installed and the CAM.R can be perform without any problem).

After then, I switch to Rscript for implementation. I save the data matrix and estimated DAG matrix. Then do: pruned_dag <- pruning(dataset, dag, pruneMethod = selGam, pruneMethodPars = list(cutOffPVal = 0.001, numBasisFcts = 10), output=TRUE).

However, it returns:
pruning variable: 1
considered parents:
pruning variable: 2
considered parents:
pruning variable: 3
considered parents:
pruning variable: 4
considered parents:
pruning variable: 5
considered parents:
pruning variable: 6
considered parents:
pruning variable: 7
considered parents:
pruning variable: 8
considered parents:
pruning variable: 9
considered parents:
pruning variable: 10
considered parents:

which means it doesn't consider any parents nodes. Any idea about it? Very much appreciate for any reply if possible.

hi,
thank you for making this code available. I am trying to install this without singularity, but getting stuck on the cppext package. Can you tell me the source of this? I am not sure which cppext was used. I am not able to find it in conda-forge or pytorch.
thanks,
Sushmita

Hi there,

I follow the instruction you leave, install Singularity and don't know how to run your "container.simg". I'm sure many people will have the same confusion... I just very new to something called "container" and "Singularity", really have no idea about how to run it. I spend hours and have no way due to my stupid.

Can you please have some further detailed instruction?

Regard