testing with extranious data splits #121
Comments
lubbersnick
added a commit
to lubbersnick/hippynn
that referenced
this issue
Jan 13, 2025
lubbersnick
added a commit
that referenced
this issue
Jan 22, 2025
* Update example * update imports to base package * add atomization torch code * Improve error message * fix type check in database * allow bias=None with hierarchical pretraining * add core code for atomization consistency * allow value nodes to stay stored as arbitrary python objects * Improve fallback handling for unique node finder * update interface to ase and lammps for atomization consistent networks * add test for atomization consistent learning * fix issue #121 * fix stray debugging code * update changelog and documentation * update example with additional arguments * recover if lammps installation was not complete * update default for example * add example readme * update example defaults * Fix bug with custom kernel warnings
|
Addressed in #123 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
When training completes, testing will occur over all data splits. This will include data from splits like 'failed_std_fac_forces'. Error handling needs to be added when testing such splits at the end of training so that failing to evaluate on those splits does not cause a full crash. Edits should be made around hippynn/experiment/routines.py line 362
Traceback (most recent call last):
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/tools.py", line 53, in log_terminal
yield file
File "/users/bnebgen/energeticMaterials/retrain-DFT-01/HIPNN-HC-hypSearch-0.py", line 233, in
setup_and_train(
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/experiment/routines.py", line 138, in setup_and_train
return train_model(
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/experiment/routines.py", line 322, in train_model
test_model(
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/experiment/routines.py", line 362, in test_model
evaluation_data = collections.OrderedDict(
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/experiment/routines.py", line 364, in
(key, database.make_generator(key, "eval", batch_size))
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/databases/database.py", line 446, in make_generator
data = [self.splits[split_name][k] for k in self.var_list]
File "/usr/projects/ml4chem/envs/p311a-venado-alf/lib/python3.10/site-packages/hippynn/databases/database.py", line 446, in
data = [self.splits[split_name][k] for k in self.var_list]
KeyError: 'energyperatom'
The text was updated successfully, but these errors were encountered: