search

mihirkatare / DeepMEM

Deep Learning Implementations for Sustainable Matrix Element Method Calculations [ IRIS-HEP Fellowship ]
Apache License 2.0
2 stars 0 forks source link

feat: Add neural network models #12

Closed mihirkatare closed 3 years ago

mihirkatare commented 3 years ago

Added major changes to the dataloder module. Added Neural Network for training, Utilities for loading validation/testing datasets. Robust input options.

Suggested squash and merge

* Extend hybridMT dataloader
* Add neural nets in model.py
* Add steering interface through main.py
* Add utils for validation and testing
* Extend CLI options
* Add split and learning rate hyperparameters to JSON config
* Add image files to .gitignore
matthewfeickert commented 3 years ago

@mihirkatare I saw that you were trying to bring most of the commits from PR #10 with you into this PR, and that you actually only had one new commit. So I just did an interactive rebase on this branch which dropped all the old and kept the new commit and force pushed this. Before we go further on this we should validate that this is the right behavior in the morning and then we can review.

matthewfeickert commented 3 years ago

Also what are the .pth files for here? Or perhaps a better way of saying that is "I know what .pth files are in Python, but why are the needed for this PR?"

matthewfeickert commented 3 years ago

Also what are the .pth files for here? Or perhaps a better way of saying that is "I know what .pth files are in Python, but why are the needed for this PR?"

Oh I wasn't thinking. These are the PyTorch models. Duh.

A common PyTorch convention is to save models using either a .pt or .pth file extension.