* Add graph construction for PotentialNet

* ignore jupyter notebooks

* Fix graph constrution

Set "remove_hs=True", to make sure the node numbers are consistent for diferent graphs.
Fix dataset download overwrite.

* Fix model

etypes for stage 1 are now sucessfully implemented

* Attempt to incorporate PotentialNet into binding_affnity_prediction/main.py

To fix: loss function error; args in configure.py

* update deprecated functions

* Handle batched graph

Now it's able to train by calling main.py

* Handles etypes with torch

* add device info to etypes

* Update args and fix for eval_epoch

to be tested on gpu

* Set explicit_hydrogen in mol_to_graph to None

No longer apply snitize and remove_hs when loading from PDBBind
Also updated deprecated functions

* Hyperparameters update & minor fixes

* Add stage 2 etypes

Digitize the edge distances [0.0, 2.0, 3.5, 4.0, 5.0] into 4 type of edges in stage 2

* Customized Gated_Graph_Conv for stage 1; Customize distance bins for stage 2 edge types

* Add validation for main.py; Add refined_random configure

* stage 3 update: customized hidden dimension

* reduce features and add hyperparameters search

* Add bond types to stage 2

* add bond types and bonds (from bigraph) to stage 2; add max_neighbor as hyperparameter

* Set core to be the test set; eval on each epoch; hyperparameter update

* add eval (test and val) at each epoch; save the scores at each epoch to file

* attempt to test for scaffold split, need to be fixed

* Parallel graph construction; add stratified and scaffold split for potentialnet

* Fix scaffold split; Attempt to fix multiprocessing on Linux

* Fix Tensor.nonzero() DeprecationWarning

* Add PDBBind v2007; Reorganize pdbbind.py; Write docstrings

* intial run on v2007

* add env specs

* add new agrusment whether to remove core set from refined set

* Update python/dgllife/data/pdbbind.py

Co-authored-by: jcyang <jincai.yang@foxmail.com>

* Update python/dgllife/data/pdbbind.py

Co-authored-by: jcyang <jincai.yang@foxmail.com>

* Update pdbbind.py

* Add argument for trials; add argument for saving r2

* Add save_path; fix trial number

* Add agglomerative sequence and structure split; Change PDBBind2007 download source to DGL data

* small typo fix

* small_typo_fixed

* DocStrings

* Improvement for handling arguments; Update configure.py

* Set default explicit_hydrogens=False; cleanup yuhang_logs

* Add README; Increase batch size; Add pin_memory

* Update README

* Update README.md

* Fix typos; Change best test_epoch to best val_epoch for reporting performance

* Fix typos

* Allow for customized dataset input

* Fix code formatting

* Improve formatting; Add docstrings; Add argument for num_workers

* minor fix PN graph construct

* Update configure.py; Add argument for num_workers; Improve formatting.

* Update README

* Fix typos and formatting

* Fix typos and formatting

* minor fix

* minor fix

* minor fix

* Customized featurizer

* Update num_workers argument

* Ignore uncommon node features

Ignored features:
atom Cs, Sr;
implicit valence 4;
total degree 5,6,7,8;

* change hyperparameters in configure.py

* Update performance with new parameters

* Update performance in README

Co-authored-by: Mufei Li <mufeili1996@gmail.com>
Co-authored-by: Ubuntu <ubuntu@ip-172-31-29-176.us-west-2.compute.internal>
Co-authored-by: Zhuhui-node108 <zhuhui@nibs.ac.cn>
Co-authored-by: zhuhui-in <69226545+zhuhui-in@users.noreply.github.com>
Co-authored-by: jcyang <jincai.yang@foxmail.com>