Failing to reproduce the code

[paulagd]

Hi,

I'm trying to run your code and i'm getting an error on the evaluation function. Could you please tell me what am I doing wrong?

    > CUDA_VISIBLE_DEVICES=1 python run_ngcf.py --dataset ml-100k
    n_users=943, n_items=1682
    n_interactions=100000
    n_train=80064, n_test=19936, sparsity=0.06305
    Creating interaction matrices R_train and R_test...
    Complete. Interaction matrices R_train and R_test created in 1.9346222877502441 sec
    Loaded adjacency-matrix (shape: (2625, 2625) ) in 0.051796674728393555 sec.
    Initializing weights...
    Weights initialized.
    Start at 2021-02-18 18:04:15.554979
    Using cuda for computations
    Params on CUDA: True
    Epoch: 0, Training time: 8.75s, Loss: 30.0306
    Traceback (most recent call last):
      File "run_ngcf.py", line 107, in <module>
        k)
      File "/home/pgd721/work/NGCF_pytorch/utils/helper_functions.py", line 142, in eval_model
        pred_items.scatter_(dim=1, index=test_indices, src=torch.tensor(1.0).cuda())
    RuntimeError: Index tensor must have the same number of dimensions as src tensor

Thank you very much!

[goohanjun]

line 137, 140 must be
pred_items.scatter_(dim=1, index=test_indices,src=torch.ones_like(test_indices).float().cuda())
topk_preds.scatter_(dim=1, index=test_indices[:, :k], src=torch.ones_like(test_indices[:, :k]).float().cuda())

[Lim-Sung-Jun]

line 137, 140 must be pred_items.scatter_(dim=1, index=test_indices,src=torch.ones_like(test_indices).float().cuda()) topk_preds.scatter_(dim=1, index=test_indices[:, :k], src=torch.ones_like(test_indices[:, :k]).float().cuda())

thanks a lot!