How to train with conservation (Add parameter -conservation_dir_train)
skodapetr opened this issue · comments
Commands such as
./prank.sh eval-predict ../p2rank-datasets/coach420.ds \
-c distro/config/conservation \
-conservation_dir 'coach420/conservation/e5i1/scores'
can be utilized to use conservation score to evaluate performance on a dataset.
I would expect train command to be similar ie.
.\p2rank\prank.bat traineval
-t .\datasets\chen11.ds \
-e .\datasets\joined(mlig).ds \
-threads 4 -rf_trees 128 -fr_depth 6 -delete_models 0 -loop 1 -seed 42 \
-c distro/config/conservation
-conservation_dir ...
however, it is not clear how to specify two directories as inputs for conservation i.e. one for chen11 and one for joined dataset.
Or do I have to merge all files into a single directory?
Unfortunately currently there is no way how to specify separate conservation directories for training and test datasets. I will add the -conservation_dir_train in the next version.
In the meantime as a workaround you can use one of 2 approaches:
- create a single directory with conservation scores for both datasets
- put the conservation score files to the same directory as pdb files
For the second approach - is it necessary to specify the conservation_dir ?
I will keep it open since adding -conservation_dir_train param should be implemented.
In version 2.2 -conservation_dir param was replaced with -conservation_dirs which can contain multiple directories.