Hi, I appreciate your work very much. You have reproduced many classic KT models, which is very helpful to people who study KT.

I want to quote your recurring results in my paper, especially the recurring results of DKVMN.
Do you have any papers that record the recurring results? I want to cite it. ^_^

Hi seewoo,
I wonder why EdNet-KT1's QUESTION_NUM(at constant.py) is 18,143.
cus as I refer to official EdNet Dataset github link, I found that its total number of problems is 13,169.
How come your EdNet-KT1 problem number grows?

Thank you.

First, Thank you for implementation and results,
I want to experiment again with your implementation, but dataset link is dead.
Can you please update the link if possible?

Thank you.

Hi Seewoo
I find that there are multiple tags assigned to question meaning related to different lectures.
In the input how are are you taking the tag if there are multiple tags for a question..

Hello,your datasets link is invalid,can you resend it again?

Hello seewoo5,
thanks for sharing your implementation of the DKVMN model.
So if i want to change the model so it predicts the response for all question and not only the last one, do i have to change the code like this?

 def forward(self, question_id, interaction, seq_len=MAX_SEQ):
        predictions = []
        value_memory=self.init_value_memory()
        t=0
        while (t<seq_len-1):
            interaction_emb=self.interaction_emb(interaction[:,t])
            question_emb=self.question_emb(question_id[:,t])
            read_content= self.read(question_emb, self.key_memory, value_memory)
            
            out=torch.cat((read_content, question_emb), dim=-1)
            out=self.linear(out)
            out=self.tanh(out)
            out=self.linear2(out)
            predictions.append(out)
            value_memory=self.write(interaction_emb, question_emb, self.key_memory, value_memory)
            t += 1

        output=torch.cat((predictions), dim=1)
        return output

Hello @seewoo5,

Thanks for sharing your implementations. I checked your code on DKT and DKVMN. It seems your models predict only one target_id for a single sequence. Am I right?

Thanks.
Chunpai

1、there not exists pre-preocessed dataset of ASSISTments2009
2、no '5 subdirectories ',just find them all together
3、the users' csv file is not in place as the code want
Could you please update the given download link or tell me what to do?