What's Changed
- [BUGFIX] Support MRSTFT on Apple Silicon by @sdatkinson in #568
- [FEATURE] init_from_nam for LSTM and Linear architectures by @sdatkinson in #569
- [BUGFIX] Don't ignore sample rate in initialization from .nam file by @sdatkinson in #571
- [FEAUTRE]
.teardown()method for datasets by @sdatkinson in #573 - [BUGFIX]
nam.model.init_from_nam(): Support files without"sample_rate"key. by @sdatkinson in #574 - [nit] Fix a few minor typos by @xconverge in #575
- [BUGFIX] Do not fail to compensate recording latency if manual latency is provided by @sdatkinson in #578
- [FEATURE] Define handshakes between datasets and NAMs by @sdatkinson in #580
- Refactor out loss dict creation by @sdatkinson in #583
- Add a pointer to the main website by @mauriciopoppe in #584
- [FEATURE] Custom losses by @sdatkinson in #587
- Fill in model parameter details in Colab cloud training section by @ehurturk in #585
- Update documentation on PyTorch install for computers with an nVIDIA GPU by @sdatkinson in #590
New Contributors
- @xconverge made their first contribution in #575
- @mauriciopoppe made their first contribution in #584
- @ehurturk made their first contribution in #585
Full Changelog: v0.12.0...v0.12.1