some strange bugs in the spareKT class

[figurexyang]

Hello, I've found some strange bugs:
Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

[anorigin]

Hello, I've found some strange bugs: Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

I have the same problem as you. Have you solved it?

[figurexyang]

Hello, I've found some strange bugs: Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

I have the same problem as you. Have you solved it?

Not yet, I've found other bugs that cause it to fail

[sonyawong]

Hello, I've found some strange bugs: Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

I have the same problem as you. Have you solved it?

Not yet, I've found other bugs that cause it to fail

Hi, thank you for your interest in our work. How did you set the emb_type of your run command? In sparseKT, the emb_types of "sparseKT-topk" and "sparseKT-soft" are "qid_sparseattn" and "qid_accumulative_attn" respectively. As the emb_type contains the string "attn", "preds" and "c_reg_loss" will be defined by lines 244-246.

[figurexyang]

Hello, I've found some strange bugs: Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

I have the same problem as you. Have you solved it?

Not yet, I've found other bugs that cause it to fail

Hi, thank you for your interest in our work. How did you set the emb_type of your run command? In sparseKT, the emb_types of "sparseKT-topk" and "sparseKT-soft" are "qid_sparseattn" and "qid_accumulative_attn" respectively. As the emb_type contains the string "attn", "preds" and "c_reg_loss" will be defined by lines 244-246.

Hi，Thank you very much for your help, and I apologize for the delay in my response. After you provided your suggestions, I set up the command line like this: CUDA_VISIBLE_DEVICES=0 nohup python wandb_sparsekt_train.py --dataset_name=assist2015 --use_wandb=0 --add_uuid=0 --emb_type="qid_sparseattn".

I suggest changing the original line 8 in example/wandb_sparsekt_train.py from parser.add_argument("--emb_type", type=str, default="qid") to parser.add_argument("--emb_type", type=str, default="qid_sparseattn"), as it might make your project easier to understand. Your work is truly inspiring. Thank you again!

[sonyawong]

Hello, I've found some strange bugs: Am I understanding this correctly? It seems that the two variables, namely preds and c_reg_loss, which are finally returned by the forward() method in the spareKT class, haven't been defined anywhere before being returned. Nor have they been found among the global variables. Should these two variables, preds and c_reg_loss, be defined before this point? Thank you for your help.

I have the same problem as you. Have you solved it?

Not yet, I've found other bugs that cause it to fail

Hi, thank you for your interest in our work. How did you set the emb_type of your run command? In sparseKT, the emb_types of "sparseKT-topk" and "sparseKT-soft" are "qid_sparseattn" and "qid_accumulative_attn" respectively. As the emb_type contains the string "attn", "preds" and "c_reg_loss" will be defined by lines 244-246.

Hi，Thank you very much for your help, and I apologize for the delay in my response. After you provided your suggestions, I set up the command line like this: CUDA_VISIBLE_DEVICES=0 nohup python wandb_sparsekt_train.py --dataset_name=assist2015 --use_wandb=0 --add_uuid=0 --emb_type="qid_sparseattn".

I suggest changing the original line 8 in example/wandb_sparsekt_train.py from parser.add_argument("--emb_type", type=str, default="qid") to parser.add_argument("--emb_type", type=str, default="qid_sparseattn"), as it might make your project easier to understand. Your work is truly inspiring. Thank you again!

Hi, we have refined the example/wandb_sparsekt_train.py. Thank you again for your suggestion and interest!