Bug fixes & updates

[ben-nowacki]

Description

Collections of small fixes and feature improvements.

1. Added support of Experiment file serialization that omits the raw FeatureSet data. Previously, add data contained in FeatureSet was forced into the final Experiment.mml file. This is prohibitive for large datasets. Now, experiments can be saved without FeatureSets, where FeatureSets are re-merged on Experiment.load(). The Experiment file still records all metadata of the prior FeatureSet and validates the re-attached FeatureSets.
2. Fixed issue where serialized result containers (eg CVResults) stored only the node ID of the node that produces some given results. This makes it very hard to determine which results belong to which node when there is no active ExperimentContext. The mapping of node ID to node label is now preserved in result containers. Users can now query results by node ID or node label, even when there is no active experiment context.
3. Minor docstring and convenience method updates:
   • Accelerator class has new method to list all available hardware for the active device
   • repr method added to ModelGraph
   • SequentialCNN default padding size is now relative to kernel size
4. New TemporalCNN class added (only implemented in PyTorch). Uses residual dilated causal convolutions. Based on: https://github.com/locuslab/TCN

Related Issues

N/A

How Has This Been Tested?

• Local pytest run
• CI passes
• Manual checks (describe if needed)

Checklist

• My code follows project style guidelines: nox -s pre-commit
• I have added tests that prove my fix is effective or my feature works: nox -s unit
• I have updated documentation if needed: nox -s docs
• I have linked related issues

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 22.67442% with 399 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
modularml/models/torch/temporal_cnn.py	18.34%	89 Missing ⚠️
modularml/core/io/handlers/experiment_handler.py	3.57%	54 Missing ⚠️
modularml/core/experiment/results/phase_results.py	15.38%	44 Missing ⚠️
modularml/core/data/sample_collection.py	6.25%	30 Missing ⚠️
modularml/core/training/loss_record.py	14.28%	24 Missing ⚠️
modularml/core/references/experiment_reference.py	20.00%	20 Missing ⚠️
modularml/core/topology/model_graph.py	20.83%	19 Missing ⚠️
modularml/core/experiment/results/train_results.py	28.57%	15 Missing ⚠️
modularml/utils/nn/accelerator.py	11.76%	15 Missing ⚠️
modularml/core/experiment/experiment.py	25.00%	12 Missing ⚠️
... and 17 more

📢 Thoughts on this report? Let us know!