Skip to content

Fix temporal mismatch and data module robustness#1

Open
snuconnectome wants to merge 5 commits intomainfrom
fix/temporal-mismatch-and-data-module-refactor
Open

Fix temporal mismatch and data module robustness#1
snuconnectome wants to merge 5 commits intomainfrom
fix/temporal-mismatch-and-data-module-refactor

Conversation

@snuconnectome
Copy link
Copy Markdown
Collaborator

Summary

This PR addresses two main issues:

  1. TypeError in DataModule: Refactored to only calculate when needed. This prevents a when is (common in dry runs or when using the dataset).
  2. Temporal Dimension Mismatch: Synchronized the temporal component of with in . This ensures consistency between the data loader and the model's positional embeddings, resolving a during training initialization.

Jiook Cha and others added 5 commits January 4, 2026 22:01
- Add Music and Narratives dataset classes to datasets.py
- Update data_module.py to support new dataset types
- Add preprocess_music.py for music dataset preprocessing
- Add train_foundation.py for foundation model training
- Add TitansNeuro model with subject conditioning capability
- Update dataset split structure for pretraining workflows
- Remove obsolete dummy split files

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Sonnet 4 <[email protected]>
b43cc15 Update on Overleaf.

git-subtree-dir: paper
git-subtree-split: b43cc150ea6ff6e4f64fab75f173381242cb1d5a
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant