New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Model not training #19
Comments
Hi @Thomas191, I just tried recreating your directory structure on my system and it worked fine. The predictions should definitely be placed in the highest numbered folder. Otherwise you're making predictions from an untrained model which will be junk. Training a model from scratch is also quite time-intensive. I noticed that you don't have a gpu assigned. Did you try an option like |
Hi! Input file contains >1 alignments, but UCSC A2M formatted output file can only contain 1 Traceback (most recent call last): Again, any help is much appreciated. |
Yes unfortunately it doesn't support multiple chains at the moment. You'd have to input them separately. |
@Thomas191 how did you convert the tertiary file to a pdb? I have my output but have no idea how to interpret it. |
@Thomas191 can you mention how exactly command structure that solve this problem for you ? |
Hello! I have been attempting to run this code for a couple of weeks but seem to have hit a dead end.
I am running the model on Ubuntu 18.04, with Tensorflow GPU installed (and verified with other code) and with CUDA 10.0 and CuDNN 7.6.1.
My end goal is to use CASP12 to predict the structure of around 1000 proteins.
At the moment I am using CASP10 (to save space) and trying to predict the structure of just one sequence to test the model.
Here is my folder structure:
WD/hmmer-3.2.1
WD/rgn/data_processing/
WD/rgn/model/
WD/proteinnet10
WD/RGN10/data/ProteinNet10Thinning90/testing
WD/RGN10/data/ProteinNet10Thinning90/training
WD/RGN10/data/ProteinNet10Thinning90/validation
WD/RGN10/runs/CASP10/ProteinNet10Thinning90/1
WD/RGN10/runs/CASP10/ProteinNet10Thinning90/2
...
WD/RGN10/runs/CASP10/ProteinNet10Thinning90/logs
WD/RGN10/runs/CASP10/ProteinNet10Thinning90/checkpoints
WD/RGN10/runs/CASP10/ProteinNet10Thinning90/configuration
WD/RGN10/logs
This is the last line of code:
rgn/model/protling.py RGN10/runs/CASP10/ProteinNet10Thinning90/configuration -d RGN10 -p -e weighted_testing
When the model runs there doesn't appear to be any errors, however the prediction is placed in folder number 1 and not in the highest number folder as would be expected.
Following the comments in another issue, I have tried deleting all the numbered folders and just training the model using the following code:
rgn/model/protling.py RGN10/runs/CASP10/ProteinNet10Thinning90/configuration -d RGN10
This only creates folder 1, logs, and checkpoints folders.
Likewise for the following code:
rgn/model/protling.py RGN10/runs/CASP10/ProteinNet10Thinning90/configuration -d RGN10 -p -e weighted_testing
Where once again only folder 1, logs, and checkpoints folders are created and the prediction for our sequence is placed in folder 1.
We have looked at this prediction and have converted it to a PDB file to view in PyMol, however the output is a helical structure (completely different to what a folded protein would look like).
We would appreciate any suggestions you have about how to fix this issue.
The text was updated successfully, but these errors were encountered: