MannLabs / alphapeptdeep

Deep learning framework for proteomics
Apache License 2.0
108 stars 20 forks source link

Error in 'peptdeep --install-model' #133

Closed hxxhust163 closed 8 months ago

hxxhust163 commented 8 months ago

Hi

Sorry to bother you. I have peptdeep installed in my linux system. when I ran the command 'peptdeep library settings_yaml'. I got an error like this. Screenshot from 2024-01-22 20-38-53

So, I downloaded the 'pretrained_models.zip' manually, and then I ran the command "peptdeep install-models --model-file ./pretrained_models.zip --overwrite True" , I got the same error.

All I did was followed the instructions. Please help to fix the problem, thanks very much!

Best Xiaoxiang

hxxhust163 commented 8 months ago

Oh, I fixed this problem by moving the 'pretrained_models.zip' into my /home directory. Thanks! But I was in trouble with other problems. So, I am writing to ask for help.

I installed the latest version of peptdeep (1.1.4) and in using peptdeep, I am encountered with two questions.

  1. In library prediction using precursor table, I compared the results of two different models, 'generic' and 'hla', which can be selected in the setting.yaml file. But the results of these two model are the same. So, I checked the 'pretrained_models.zip', but found no hla model. It looks like this: Screenshot from 2024-01-24 16-45-59

I wonder where I can download the hla pretrained model?

  1. Also during the library prediction using precursor table, I ran into two small problems in peptdeep-1.1.4. One is the keyerror and another one is the warning. Looks like bellow:

Screenshot from 2024-01-24 16-57-31

But it was successful in peptdeep-1.0.2 without the error and warning when using the same input precursor table. The precursor table is in csv format and looks like this: Screenshot from 2024-01-24 17-04-53

I also installed the peptdeep-1.1.4 in windows with python3.9. And got the same keyerror and warning.

Please help to figure those out. Thanks very much!

Best wishes Xiaoxiang

jalew188 commented 8 months ago

Yes, HLA and generic models are not trained trained togethor so they are identical.

For KeyError issues, I will test on my own computer

hxxhust163 commented 8 months ago

OK, thanks! Looking forward for your reply. Maybe there is something wrong in my operation.

Happy snowing! [laugh]