[WIP] Multimodal - Githubissues

YutackPark commented 3 months ago

Running example_inputs/training/input.yaml fails with below error. Torch version 2.4.0

Model building...
Traceback (most recent call last):
File "/home/parkyutack/.local/venv11/svn_dev/bin/sevenn", line 8, in <module>
sys.exit(main())
^^^^^^
File "/home/parkyutack/.local/venv11/svn_dev/lib/python3.11/site-packages/sevenn/main/sevenn.py", line 85, in main
train(global_config, working_dir)
File "/home/parkyutack/.local/venv11/svn_dev/lib/python3.11/site-packages/sevenn/scripts/train.py", line 70, in train
model = build_E3_equivariant_model(config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/parkyutack/.local/venv11/svn_dev/lib/python3.11/site-packages/sevenn/model_build.py", line 401, in build_E3_equivariant_model
shift = torch.FloatTensor(shift)
^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: new(): data must be a sequence (got float)

YutackPark commented 3 months ago

Line number < 80 for processing_dataset.py (+refactoring would heplful)

YutackPark commented 2 weeks ago

[ ] sevenn_cp tool that summarizes checkpoint + append/assign modality to the checkpoint
[ ] LAMMPS + mm
[ ] test code for multimodal
[ ] test code for multimodal + LAMMPS
[ ] raise warnings for not-compatible checkpoints continue
[ ] SevenNet-MF
[ ] multimodal presets
[ ] Minimal documentation/guide for multimodal training + usage (+ paper in README.md)

MDIL-SNU / SevenNet

[WIP] Multimodal #71