torchmd-net
torchmd-net copied to clipboard
torchmd
Neural network potentials
The argument parsing mechanism assumes that arguments (CLI or in the yaml file) will be processed in always the same order, but Argparse offers no such guarantee. ```python parser.add_argument('--load-model', action=LoadFromCheckpoint,...
Help~ When I using ` from torchmdnet.models.model import load_model ` showing that 'ImportError: Could not find module 'torchmdnet_extensions' in torchmd-net-main\torchmdnet\extensions'
We currently use `CSVLogger` from `torch.lightning` to write metrics to a CSV file in the log directory. However, when using `load_model` (i.e., loading from a checkpoint), or more in general...
I believe the MACEOFF dataset contains total energies, with the provided TensorNet-MACE.yaml it will not train properly. The losses diverge. It needs to be run with `remove_ref_energy: true` and a...
Currently using the training argument "output_mlp_num_layers" will not actually do anything with a scalar output model. The keyword argument here needs to be changed to "num_hidden_layers": https://github.com/torchmd/torchmd-net/blob/f6c0c16932e59373b631ceca686d810832a66b5d/torchmdnet/models/output_modules.py#L79 To correspond to...
This PR introduces a custom version of MDCATH while preserving the original functionality to ensure backward compatibility. New Arguments: - `source_file `and `file_basename`: these args enable users to create and...
We are training a model and require a plot of loss function with epochs while the model is training. How can we do this with torchmd?
When reading the TensorNet paper (_TensorNet: Cartesian Tensor Representations for Efficient Learning of Molecular Potentials_), specifically in section A.3, the authors describe the details of training on scalar, vector, and...
We are working on Chignolin Coarse Grain Tutorial from this repository: https://github.com/torchmd/torchmd-cg.git We are working on a Machine Learning project pertaining to Coarse-graining of protein molecules. As a part of...