Tags

1) Feature_enhance is added
2) model_free_locscale by default is using the predicted model maps
3) emmernet models are trained against hybrid and atomic model maps (60k dataset, 15 epochs, 0.0001 lr)
4) Filtered cubing for faster prediction (using both CPU and GPU)
5) Logging with loguru

1) Pseudo-model elements now back to oxygen and residues HOH
2) Lower bound threshold calculated using GMM (gaussian kde) 
3) Removed "no_mask" for model based and hybrid
4) Correction factor for num_atoms (1.55)

1) proper element composition for pseudo-models
2) ADP shift determined by 1% CDF
3) Fixed mess in locscale report

Locscale version used to generate the "epsilon" version of the training targets for Emmernet.

Locscale version used to generate the "delta" version of the training targets for Emmernet.

The latest update for LocScale includes: (1) Default refmac iterations of 10 (2) varying radius for atomic model mask generation for hybrid locscale

The latest update for LocScale includes: (1) Integrated pseudo-atomic model building, (2) Set the averaging filter window for difference mask calculation to 3 pixels (3) Suppressed print statements

LocScale version used for generating the training dataset for EMmerNet

LocScale version used for generating the training dataset for EMmerNet

This is a pre-release tag before version 2 deployment

version 0.1