SpeechBrain 1.0 #2435
mravanelli
announced in
Announcements
SpeechBrain 1.0
#2435
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Dear Users,
After a year of hard work, we are thrilled to announce the official release of SpeechBrain 1.0 🚀. This milestone comes with numerous enhancements and advancements.
You can explore a comprehensive summary of our improvements here: SpeechBrain 1.0 Summary.
Among our achievements, we have made significant improvements in speech recognition, enhancing search functionalities through integration with K2 for finite-state transducers, CTC decoding, and n-gram rescoring. Additionally, we have introduced novel models such as Streamable Conformer Transducers, Branchformers, and Hyper-conformer, among others, to improve performance and speed.
Furthermore, SpeechBrain now supports a broader spectrum of tasks for speech,audio, text, and EEG processing. We improved our integration with HF models to make importing any model from HF easier. We implemented modern techniques and models, including continual learning, diffusion models, hyper-networks, Bayesian ASR, and more.
You can now easily use large-language models(LLMs) and fine-tune them with our data, or simply employ them for rescoring ASR hypotheses.
We have created a new benchmark repository featuring useful benchmarks for self-supervised learning (MP3S), EEG processing (SpeechBrain-MOABB), and continual learning of new languages (CL-MASR).
We hope to have made a meaningful contribution to our community, and you are welcome to share your feedback with us.
Finally, if you value our commitment to the community, please consider starring our project on GitHub ⭐.
Stay tuned for the future because we have big plans ahead.
Thank you for being part of SpeechBrain!
Regards,
Mirco
Beta Was this translation helpful? Give feedback.
All reactions