Thunder-speech

Fix
* Improve typing on decoder and optimizer arguments ([`83c0758`](https://github.com/scart97/thunder-speech/commit/83c075897f7c38f92653e941860f40e8eb3d1e88))

Fix
* **huggingface:** Correct loading of checkpoints that have unused tokens ([`887608e`](https://github.com/scart97/thunder-speech/commit/887608e7f41ef47d71429ddb9211cbc4eb69d581))
* **huggingface:** Fix loading of pretrained only models, that lack the tokenizer ([`65106ee`](https://github.com/scart97/thunder-speech/commit/65106eea7dd6a4dc7c4f13b6a6d74567835b306e))

Fix
* Trigger release with updated dependencies ([`3b59ffe`](https://github.com/scart97/thunder-speech/commit/3b59ffe446183ccefb8229eebaca77a4e5e098df))

Documentation
* **readme:** Update readme dev instructions ([`6e39bd9`](https://github.com/scart97/thunder-speech/commit/6e39bd9d99d61f7ec0a07fd39b732b7b17593c8e))