model.pth not found

[joaomcteixeira]

Dear @promethiume

I have recently found your article and software, and I am trying to use it.

I think I have managed to install everything required, but when running python Code/main.py --id ### I eventually get this message:

FileNotFoundError: [Errno 2] No such file or directory: 'model.pth'

Am I missing some file? I have never worked with AI models before; sorry if I am making a naive question.

Best and congratulations on such a nice article!

[joaomcteixeira]

Hi,

I could overcome the issue by commenting these lines:

LLPSWise/Code/main.py

Lines 72 to 74 in bf55c96

	model=torch.load("model.pth")
	model.to(device)
	model.eval()

As the model variable was defined twice. Hence, remaining only with:

LLPSWise/Code/main.py

Line 71 in bf55c96

model = ANNmodel(input_dim, hidden_dim, output_dim)

But I don't know if this is the correct solution.

Best,

[promethiume]

Thanks for your interest. We have a trouble on uploading our model file to GitHub due to the size limitation.
It won't work by just commenting those lines. It will cause a "low propensity" result.
I've just finished the network calculation on all the uniprot protein and currently uploading all the results file. I'll let you know after I finished it and I'll update the README file for the result visualization.

Besides, we are working on a web server including some new features and a brand new lips prediction algorithm. Hopefully it will coming soon.

[joaomcteixeira]

Hi,

Thanks so much for your reply. Yes, we are very interested in your software, and I have been trying it in many ways these days but without successful results, as you know. I am looking forward to your updates on the calculations for the whole uniprot.

Also having a webserver would be great. 👍

Regarding large files on GitHub, you can try their service for large files https://git-lfs.github.com/

On another note, I was playing a bit in my fork and wrote there some hints on the local installation until biopython deploys and a small trick to download uniprotID upon request instead of having all downloaded beforehand.

1. 1c4fcae
2. 9209d9a

I will pay attention to your updates,
Cheers,

[promethiume]

I've uploaded our result on templates.tar.gz
You can download it, and use the app.py to check the network result (see usages on README).

I'll have a look at the large file service and the uniport issue later.

Thanks.

[joaomcteixeira]

Thanks so much, I will have a look to it now.

The UniProt reading issue happens because the BioPython code in PyPI is from June last year. This summer, UniProt files were updated with additional features which made the SeqIO broken. Recently, people in BioPython solved it. So, you need to install biopython from the source instead of PyPI. But that conflicts with the bio package. To solve all this dependency twist, I wrote some install instructions that are in my fork 👉 https://github.com/joaomcteixeira/LLPSWise/tree/jmain#installation. Likely, you haven't faced this issue because you are using the BIOGRID and UniprotDB from before summer.

[joaomcteixeira]

I've uploaded our result on templates.tar.gz

They work, thanks so much. It's not actually needed to launch the flask app. Opening the html files with the browser gives the same result.

Many thanks; looking forward to updates!