[WIP][Benchmark] SSL Baselines (MAE, SimCLR, LeJEPA, BarlowTwins, NNCLR, Dino 1)

[sami-bg]

What does this PR do?

Here are the instructions for submitting SSL benchmark runs to SLURM:

  SSL Benchmark SLURM Submission

  # Activate the virtual environment
  source /oscar/home/sboughan/stable-datasets/.venv/bin/activate

  # Single dataset, single model
  SLURM_PARTITION=3090-gcondo python -m stable_datasets.benchmarks.self_supervised.main \
      --multirun --config-name slurm 'hydra.launcher.qos=null' \
      dataset=cifar10 model=simclr backbone=vit_small

  # Multiple datasets, single model
  SLURM_PARTITION=3090-gcondo python -m stable_datasets.benchmarks.self_supervised.main \
      --multirun --config-name slurm 'hydra.launcher.qos=null' \
      dataset=cifar10,stl10,svhn model=dino backbone=vit_small

  # Multiple models on multiple datasets (creates one job per combo)
  SLURM_PARTITION=3090-gcondo python -m stable_datasets.benchmarks.self_supervised.main \
      --multirun --config-name slurm 'hydra.launcher.qos=null' \
      dataset=cifar10,cifar100 model=simclr,dino,mae,lejepa,nnclr,barlow_twins backbone=vit_small

  # All datasets (auto-expands to every available dataset)
  SLURM_PARTITION=3090-gcondo python -m stable_datasets.benchmarks.self_supervised.main \
      --multirun --config-name slurm 'hydra.launcher.qos=null' \
      dataset=all model=simclr backbone=vit_small

  # Collect results into a LaTeX table (with supervised baselines)
  python -m stable_datasets.benchmarks.self_supervised.collect_results \
      --entity stable-ssl --project stable-datasets-benchmarks \
      --supervised-entity samibg --supervised-project stable-datasets \
      --latex results.tex

  Notes:
  - Set SLURM_PARTITION to your partition (e.g., gpu, 3090-gcondo)
  - Make sure to put the right paths on config.yaml (it's hardcoded for my path)
  - Add 'hydra.launcher.qos=null' if your partition doesn't require a QOS
  - Results are logged to W&B under stable-ssl/stable-datasets-benchmarks
  - The collect_results.py script caches finished W&B runs locally, so re-runs are fast

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue? Please add a link to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines
• Did you write any new necessary tests?

Who can review?

Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.

[Leon-Leyang]

Could you run pre-commit run --all-files so that the precommit test passes.