- Fix bug in getting the watermark with non-empty message created in CPU, while the model is loaded in CUDA - Update Fix bug in building the model card programmatically (not via .YAML file using OmegaConf) - Add support for HuggingFace Hub, now we can load the model from HF. Unit tests are updated
0.1.2
- Add py.typed to make audioseal mypy-friendly - Add the option to resample the input audio's sample rate to the expected sample rate of the model (https://github.com/facebookresearch/audioseal/pull/18) - Move `attacks.py` to non-core code base of audioseal - Remove duplicate module `SEANetEncoderKeepDimension` in `audioseal.lib.audiocraft.modules.seanet` and `audioseal.models`
0.1.1
- Fix [issue](https://github.com/facebookresearch/audioseal/issues/7) in installing audioseal from pypi due to conflict with audiocraft package - Fix typos in example notebooks - Update checkpoint to be Windows-compatible