training-forensics
Compare a fine-tune with its base: the weights list which vocabulary tokens its training data contained. 421 release/base comparisons across 580 open models.
- 1
- Python
- MIT
- 3 d ago
Compare a fine-tune with its base: the weights list which vocabulary tokens its training data contained. 421 release/base comparisons across 580 open models.
Open-source toolkit for Fourier-based audits of trained quantum neural networks.