corochann
fixing relgcn, relgat calculation as well, due to ggnnpreprocessor change.
BaseSplitter's train_valid_split checks assert len(test_inds) == 0. when i use RandomSplitter, it failed. i guess calculation of ``` train_data_size = int(len(dataset) * frac_train) valid_data_size = int(len(dataset) * frac_valid) ``` may...
graph_conv module is better than `update` ??
TODO - addHs before EmbedMolecule - check option of EmbedMolecule - execute UFF or MMFF optimization
currently unittest is not implemented.
I think it is better to **merge in another branch** PR is to upgrade the code for chainer v2~. I tested that this code can be executed on chainer v...
[Note] this PR is optional, if you agree this is ok please merge it but it is also ok to ignore this PR. Currently we must set arguments to execute...
When user is not installed `cupy` (only CPU environment), `from chainer.cuda import cupy` will fail. After this PR, `train.py` can be executed in CPU environment without `cupy` installed.