Adding ML custom force for isotropic CG systems

[marjanalbooyeh]

This pull request focuses on the initial implementation of machine learning models for coarse-graining simulations. The changes include adding new utility functions, creating a custom force class, defining a neural network model, and providing a tutorial notebook.

Changes:
Introduced IsotropicCustomForce in flowermd/library/ml_forces.py to integrate a PyTorch model with the simulation framework.

Defined LennardJonesNeighborModel and its configuration in flowermd/modules/coarse_graining_ML/isotropic_coarse_graining.py for predicting forces based on neighbor distances.

Added functions adjust_periodic_boundary, find_neighbors, and neighbors_dr to flowermd/internal/box_neighbors_utils.py for handling periodic boundary conditions and neighbor calculations.

Added a new Jupyter notebook tutorials/6-flowermd-ML-coarse-graining.ipynb to demonstrate the usage of the pre-trained machine learning model for coarse-graining simulations.

Note: I'll wait for the units PR to be merged before merging this one.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[chrisjonesBSU]

Since we probably don't want to add pytorch as a hard depdency, anywhere we import we should check if it's installed and give an error message if it's not. Maybe it says something like

You need to have pytorch installed to use this functionality. You can install it with conda or mamba using mamba install -c conda-forge pytorch

[codecov]

Codecov Report

Attention: Patch coverage is 79.62963% with 55 lines in your changes missing coverage. Please review.

Project coverage is 93.40%. Comparing base (a600b73) to head (8b312c5).
Report is 49 commits behind head on main.

Files with missing lines	Patch %	Lines
flowermd/library/ml_forces.py	27.58%	21 Missing ⚠️
flowermd/internal/box_neighbors_utils.py	25.00%	18 Missing ⚠️
flowermd/base/simulation.py	87.71%	14 Missing ⚠️
flowermd/internal/utils.py	90.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #178      +/-   ##
==========================================
- Coverage   94.77%   93.40%   -1.37%     
==========================================
  Files          26       29       +3     
  Lines        1970     2213     +243     
==========================================
+ Hits         1867     2067     +200     
- Misses        103      146      +43     

Files with missing lines	Coverage Δ
flowermd/__init__.py	100.00% <100.00%> (ø)
flowermd/base/system.py	92.52% <100.00%> (+0.13%)	⬆️
flowermd/internal/__init__.py	100.00% <100.00%> (ø)
flowermd/internal/exceptions.py	89.74% <100.00%> (ø)
flowermd/internal/units.py	100.00% <100.00%> (ø)
flowermd/library/__init__.py	100.00% <100.00%> (ø)
flowermd/library/simulations/tensile.py	100.00% <100.00%> (ø)
...lowermd/modules/surface_wetting/surface_wetting.py	94.78% <100.00%> (+0.04%)	⬆️
flowermd/utils/utils.py	93.47% <100.00%> (+0.14%)	⬆️
flowermd/internal/utils.py	93.75% <90.00%> (-6.25%)	⬇️
... and 3 more

[marjanalbooyeh]

Since we probably don't want to add pytorch as a hard depdency, anywhere we import we should check if it's installed and give an error message if it's not. Maybe it says something like

You need to have pytorch installed to use this functionality. You can install it with conda or mamba using mamba install -c conda-forge pytorch

Yeah I agree. I'll add that condition.