torchmd-net
torchmd-net copied to clipboard
torchmd
Neural network potentials
Following the discussion in https://github.com/torchmd/torchmd-net/issues/198 this PR attempts to give TorchMD_Net the ability to return more than one output ("y") and its derivative ("neg_dy"). This PR is still a draft...
This PR makes it possible to get tensor outputs from TensorNet. Also, it implements some tests to ensure equivariance. TODO: - [ ] Write a somewhat general tensor output model
I don't face this error on another checkpoint that is trained on a different dataset. Please let me know if any more information is required. Thanks ```shell Error: RuntimeError Traceback...
I often want to resume training from a checkpoint with `--load-model`. When I do that, I don't want to lose all the information in the `log` and `metrics.csv` files. The...
I just encountered an error I've never seen before. I used the `--load-model` command line argument to resume training from a checkpoint. At first everything seemed to be working correctly,...
Adds the `--freeze-representation`, `--reset-output-model` and `--overwrite-representation` to train.py. - Freeze representation: Makes it so that the representation model weights are not trained - Reset output model: Makes it so that...
Enhance model training by enabling Exponential Moving Average (EMA) on the model parameters. To activate this feature, simply specify the `ema_prmtrs_decay` arg in the configuration file. If not set, its...
- [x] Added Swish and Mish as possible activation functions. - [x] Added a Gated Linear Unit (GLU) module ( eq. 1 here https://arxiv.org/pdf/1612.08083 ), but the activation function is...
This PR introduces a new optional term to the loss that depends on the gradient of the force. The hope here is that by penalizing large force gradients the resulting...