minatotanaka7/deit-matching-mini

A compact DeiT for Matching

Overview

This is an experimental DeiT codebase for Matching. It keeps the small setup intentionally manageable so architecture changes can be inspected before a full training run.

Expect rough edges — this is a working sketch, not a polished release.

Repository status

One Python module contains the model and a runnable example. Generated configuration lives in config.json; the recipe in training_args.json. The checkpoint is an initialization artifact and should not be read as a trained model. Scores are deliberately absent.

Quick check

Run python model.py --help for the built-in smoke test. Because this is a custom implementation, generic automatic loading APIs require an explicit adapter before use.

Training details

1. Framework: PyTorch
2. Precision: fp32
3. Init scheme: xavier
4. Hardware: a single consumer GPU is enough
5. Typical run: minutes, not hours

Default experiment recipe

Defaults: sgd optimizer, onecycle schedule. Treat these as starting points; a real run should sweep learning rate and batch size before drawing conclusions.

Limitations

1. weights are random-init, so outputs carry no signal
2. robustness, fairness and domain transfer are untested
3. the API may change without notice while the design settles

Roadmap

Next steps, in no particular order: a cleaner entry point, input-shape docs, and a small evaluation script if time permits.

Citation

@misc{acompactdeitformatch},
  title = {A compact DeiT for Matching},
  year = {{2026}},
  publisher = {{Hugging Face}},
}

Files

1. model.py — primary artifact
2. README.md — this documentation
3. config.json — architecture configuration
4. training_args.json — default experiment settings
5. model.safetensors — initialization checkpoint

License

Released under bsd-2-clause. Review the source-data terms separately when this repository is used with external datasets.