deeprank

deeprank copied to clipboard

Go through the text and polish it

NicoRenaud

Many docking models do not have contact residues. DeepRank is reporting error for these models. Could we change the word 'error', 'valueerror' to 'warning' so that the real error can...

LilySnow

cannot generate hdf5 for some complexes

17

code and data: `/projects/0/deeprank/BM5/issue147` > Create new database ./1ZM4.hdf5 > > Start creating HDF5 database: ./1ZM4.hdf5 > Creating database : 0%| | 0/1 [00:00

LilySnow

shall we disable realign_complexes()

1

Since we have a function to align PDB files, shall we disable `realign_complexes()` from hdf5? realign_complexes() has the following issues: 1. It needs pssm files as input even though the...

LilySnow

def get_feature_value() in FullPSSM.py fails on pdb files with missing CB atoms

7

Error when running `DataGenerator()` on a pdb file with missing CB atoms. Data and code are under: `/projects/0/deeprank/BM5/issue139` Error message: > ERROR: ('B', 47, 'THR') > Traceback (most recent call...

LilySnow

I am writing an issue here just in case we forget. The code is `deeprank/utils/plot_utils.py`, which calls `cal_hitrate_successrate.py`, `boxplot.R`, `hitrate_plot.R`, and `successrate_plot.R`.

LilySnow

Tests for training/test with GPU/CUDA is lacking in `test/test_learn.py`. Need to add some test cases for that.

CunliangGeng

GPU/CUDA can be used to do feature mapping in `DataGenerator.py`, but it has not been tested so far. There is test file `test/test_generate_cuda.py`. Need to check the test file and...

CunliangGeng

regression scatter plot crashes

5

--> Scatter Plot : ./out_001_split_channels_n_8/prediction_0000.png Traceback (most recent call last): File "train_001_split-channels_regression.py", line 208, in main() File "train_001_split-channels_regression.py", line 204, in main hdf5='epoch_data.hdf5' File "/nfs/home4/sgeorgie/deeprank/deeprank/learn/NeuralNet.py", line 362, in train save_model=save_model)...

sonjageorgievska

grid size is fixed when mapOnFly = false

2

Li told me to note down this, so that everyone knows: if the training uses already mapped features in 3D, then the network will use the resolution as it is...

sonjageorgievska