search

TmacMai / CoMPILE_Inductive_Knowledge_Graph

Communicative Message Passing for Inductive Relation Reasoning（AAAI-2021）
MIT License
31 stars 4 forks source link

Hyper-parameters #4

Open yzhangee opened 2 years ago

yzhangee commented 2 years ago

Hi, thank you for sharing the code for your nice work. Since CoMPILE will be used as a baseline method to compare, I want to ask about the hyperparameters you use. Does python train.py -d data work for all 12 versions or do I need to specify the hyperparameters for different datasets?

TmacMai commented 2 years ago

Hi, thank you for your attention to our work. I have lost the paras.json files. I remember that we fine-tuned the batch size hyperparameter, which has significant influence to the performence. Therefore, if the model does not perform well, please change the batch size (e.g., 16, 32, 128).     ------------------ Original ------------------ From: @.>; Date:  Wed, Sep 21, 2022 11:22 AM To: @.>; Cc: @.***>; Subject:  [TmacMai/CoMPILE_Inductive_Knowledge_Graph] Hyper-parameters (Issue #4)

 

Hi, thank you for sharing the code for your nice work. Since CoMPILE will be used as a baseline method to compare, I want to ask about the hyperparameters you use. Does python train.py -d data work for all 12 versions or do I need to specify the hyperparameters for different datasets?

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>