RNN models and LR Schedulers

[nickwimer]

No description provided.

[Copilot]

Pull Request Overview

This PR implements RNN models and learning rate schedulers for the PT-MELT library. The changes include a new RecurrentNeuralNetwork model class with various pooling strategies, enhanced training functionality with scheduler support and early stopping, and improvements to the loss functions.

• Adds comprehensive RNN implementation with LSTM/GRU/vanilla RNN support and multiple pooling strategies
• Integrates learning rate schedulers and early stopping mechanisms into the training loop
• Refactors utility functions from standalone methods to inline implementations within blocks

Reviewed Changes

Copilot reviewed 6 out of 7 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
setup.py	Updates version and adds safetensors dependency
ptmelt/nn_utils.py	Comments out utility functions (get_activation, get_initializer, get_loss_fn)
ptmelt/models.py	Adds RNN model class, scheduler support, optimizer factory, and enhanced training loop
ptmelt/losses.py	Improves MixtureDensityLoss with MSE weighting and numerical stability
ptmelt/layers.py	Adds AttentionPool layer for RNN sequence pooling
ptmelt/blocks.py	Replaces imported utility functions with local implementations

Comments suppressed due to low confidence (2)

ptmelt/models.py:1

• Remove commented-out code that has been replaced by the torch.clamp approach on the next line.

import warnings

ptmelt/models.py:1

• Remove commented-out code. The loss calculation has been refactored to handle different reduction strategies.

import warnings

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}