search

PRBonn / phenobench-baselines

Baselines of the PhenoBench Dataset
https://www.phenobench.org
22 stars 4 forks source link

About training configuration details #5

Closed chenfh21 closed 7 months ago

chenfh21 commented 8 months ago

Hi: Thank you for providing all these baselines. Now I am utilizing the baseline you provided and attempting to solve the relative issues. I have carefully read your article on Phenobench, which stated that you would provide details about training configuration in the supplement. Unfortunately， I did not find the supplement you mentioned. So I want to know if these details consistent with the parameters in the yaml file of the baseline.

jbehley commented 8 months ago

Short answer: yes, these are the configurations that we used for the results in the paper. Only the length of training (in terms of epochs or iteration) might be missing.

Long answer: The paper is currently under review and we felt that uploading a 100MB+ file with all the additional qualitative results to arxiv makes it "inconvenient" to download. The supplement will be available via our project page or the publisher website, once it's accepted. If you need the supplement now, I can provide you a link to download.

But we can also provide missing details via a github issue. Note that we also provide the checkpoints of the models and the predictions (on validation and testset) for full reproducibility.

chenfh21 commented 8 months ago

Thanks for your response.

If it's convenient, could you please provide the supplementary materials? It perhaps would be helpful for my research. I appreciate you sharing the download link.

-----原始邮件----- 发件人:"Jens Behley" @.> 发送时间:2024-02-19 00:17:26 (星期一) 收件人: PRBonn/phenobench-baselines @.> 抄送: chenfh21 @.>, Author @.> 主题: Re: [PRBonn/phenobench-baselines] About training configuration detals (Issue #5)

Short answer: yes, these are the configurations that we used for the results in the paper. Only the length of training (in terms of epochs or iteration) might be missing.

Long answer: The paper is currently under review and we felt that uploading a 100MB+ file with all the additional qualitative results to arxiv makes it "inconvenient" to download. The supplement will be available via our project page or the publisher website, once it's accepted. If you need the supplement now, I can provide you a link to download.

But we can also provide missing details via a github issue. Note that we also provide the checkpoints of the models and the predictions (on validation and testset) for full reproducibility.

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

jbehley commented 8 months ago

Please find the draft of the supplement at https://www.phenobench.org/data/phenobench_supplement_draft.pdf

Hope that helps.