Test2016 inputs

[liyujiao11]

hi,I am very interested in your research, and I am trying to reproduce your program. It went well in the early stage, but when I finished training the 18th epoch, and the 19th epoch was about to use the Test2016, there was an error report saying that I could not find xxx.inputs. It is found that there are only five subprojects (DSSP, FASTA, HHM, OPUS, PSSM) in each sequence of the Testset2016. Is it convenient for you to share all the resources of Test2016? Thanks!

[thuxugang]

Hi, thanks for your interest. Acutually, Test2016 is not neccessary since it will not affect the training process. You can just comment the relevant codes, and test the targets in it as those in Test2018, CASP and Cameo93. :)

[liyujiao11]

OK, so when I train the model and want to test the effect, I also need to use the test set, but I find that the CASP 、Test2018 in the dataset also has no inputs and labels file. So how do I test the model? thanks～😋

…

------------------ Original ------------------ From: 徐罡 ***@***.***> Date: Fri,Jul 2,2021 0:57 PM To: thuxugang/opus_tass ***@***.***> Cc: yujiaoLee ***@***.***>, Author ***@***.***> Subject: Re: [thuxugang/opus_tass] Test2016 inputs (#2)

[thuxugang]

Please use inference.py, it only requires *.pssm and *.hhm, and both of them can be found in the dataset. FYI, xxx.inputs, xxx.labels and xxx.labels_mask are the formated files for traning, xxx.dssp, xxx.angles and xxx.ss are the native results~

[liyujiao11]

ok，thanks for your guidance😋

…

------------------ Original ------------------ From: 徐罡 ***@***.***> Date: Fri,Jul 2,2021 1:42 PM To: thuxugang/opus_tass ***@***.***> Cc: yujiaoLee ***@***.***>, Author ***@***.***> Subject: Re: [thuxugang/opus_tass] Test2016 inputs (#2) Please use inference.py, it only requires *.pssm and *.hhm, and both of them can be found in the dataset. FYI, xxx.inputs, xxx.labels and xxx.labels_mask are the formated files for traning, xxx.dssp, xxx.angles and xxx.ss are the native results~ — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or unsubscribe.