Yes I really want to run this program on large-scale datasets, but I really don't know how to optimize the code on the GPU. Please give me a guide.

I've tried to set "--epochs=5 --batch-size=32" when training the model, but still got the following:

RuntimeError: CUDA out of memory. Tried to allocate 2.00 MiB (GPU 0; 10.76 GiB total capacity; 6.91 GiB already allocated; 5.81 MiB free; 6.93 GiB reserved in total by PyTorch)

I'm just a beginner. Appreciate your sincere help.

First of all, I would like to thank you for implementing the codebase in such an efficient and comprehensive way!

Just one minor issue that I have noticed that in processing.py as you are converting raw data from CSV files to answers, questions, and features, you mention in comments that in Step 4, answers need to be shifted by 1, which I think is in line with the problem definition and the paper. However, I don't think there are any shift operations been done in preparing the raw data, nor is there any operation that does this when iterating the dataloader.

So I would like to know whether there is anything that I overlooked, or this is indeed a typo in the codebase.

Thank you!

Below is the code snippet for Step 4 from the file.

    # Step 4 - Convert to a sequence per user id and shift features 1 timestep
    feature_list = []
    question_list = []
    answer_list = []
    seq_len_list = []

Hi, this code is beautiful, but it runs so slow on my Nvidia GTX 2080 Ti, taking 248 seconds for one batch (on assist2009, batch size = 128). And once I use dataset with larger num of skills, the program corrupted due to lack of GPU memory. So I wonder what kind of GPU do you use to run this model and how long does it take to train?

Hello, I don't understand how to extract the learned graph structure. I haven't found relevant information in the code

Hi, jhljx. What a brief and beautiful implement ! But , when I run the code on my mechine, the predict values of the model were nan afer few epochs . And I found that the parameters of the model were updated by nan through back propagation. I am not sure if this is caused by gradient explosion. If it is , how to solve the problem. I have tried to decrease the learning rate and batch size , but it seems not work.