Project_final_master1

TER M1 project focused on Test-Time Training and Self-Supervised Learning.

Authors

Rayane KHATIM, Mehdi AGHAEI
Universite Paris Cite

Clean Structure

Project_final_master1/
├── README.md
├── LICENSE
├── run.py
├── papers/
│   ├── 1_sun20.pdf
│   ├── 3_When_test_time_Adaptation.pdf
│   └── TER2.pdf
├── docs/
│   └── project_tracking.md
├── notebooks/
│   ├── basicUnderstanding.ipynb
│   └── choose_paper_for_TER.ipynb
├── src/
│   ├── __init__.py
│   ├── datasets.py
│   ├── test_time_training.py
│   ├── self_supervised.py
│   └── metrics.py
├── configs/
│   ├── test_time_training.yaml
│   └── self_supervised.yaml
└── data/
    ├── raw/
    ├── interim/
    └── processed/

What Each Part Means

• papers/: your TER PDFs and reference papers.
• docs/project_tracking.md: links for presentations/progress PPTs + weekly decisions.
• notebooks/: exploration notebooks and paper-choice notebook.
• src/test_time_training.py: code skeleton for test-time adaptation/training.
• src/self_supervised.py: code skeleton for self-supervised training.
• src/datasets.py: dataset metadata/registry.
• src/metrics.py: evaluation helpers (accuracy, etc.).
• configs/*.yaml: experiment parameters (model, data path, hyperparameters).
• data/raw, data/interim, data/processed: dataset lifecycle.
• run.py: single entrypoint to run either track.

Run

python run.py --task test_time_training --config configs/test_time_training.yaml
python run.py --task self_supervised --config configs/self_supervised.yaml

Short aliases are also accepted:

python run.py --task ttt --config configs/test_time_training.yaml
python run.py --task ssl --config configs/self_supervised.yaml