Description

I am just beginning to use Cornac and, while it is nice to have a module that contains everything from data loading to model training and evaluation, I often find myself lost regarding the progress of my execution. For example, I have a script very similar to the Quickstart example, and I am using a dataset similar in size and characteristics to the ML25M to train a BPR and a WMF model. Their maximum iterations are set to 100, and the verbose is True – I want to know where I am at. However, after the model is trained, there is no logging regarding the evaluation, and it takes a huge amount of time without any output.

A snippet of my code:

ss = StratifiedSplit(
    data=data,
    chrono=True,
    fmt="UIRT",
    test_size=0.2,
    rating_threshold=4.0,
    seed=RANDOM_SEED,
)

mf = MF(
    k=EMBEDDING_DIM,
    max_iter=100,
    learning_rate=0.01,
    lambda_reg=0.02,
    use_bias=True,
    seed=RANDOM_SEED,
    verbose=True,
)
wmf = WMF(
    k=EMBEDDING_DIM,
    max_iter=100,
    learning_rate=0.001,
    seed=RANDOM_SEED,
    verbose=True,
)

models = [mf, wmf]

metrics = [
    MAE(),
    RMSE(),
    Precision(k=10),
    Precision(k=20),
    Precision(k=50),
    Recall(k=10),
    Recall(k=20),
    Recall(k=50),
    NDCG(k=10),
    NDCG(k=20),
    NDCG(k=50),
    AUC(),
    MAP(),
]

cornac.Experiment(eval_method=ss, models=models, metrics=metrics, user_based=True, verbose=True).run()

The output after 8h running:

100%|██████████████████████████████████| 100/100 [03:18<00:00,  1.99s/it, loss=2152119.50]
Optimization finished!

It takes 3 minutes to train the model, but after that, in 8 hours, all I can see is my CPU being used. How I am supposed to know the ETA and if it's working correctly?

Thanks in advance.