Spaces:
Runtime error
Apply for community grant: Personal project
Hi @reach-vb and @sanchit-gandhi 👋🏻
I hope you've had an amazing week!
During the last few days, I've fine-tuned multiple Whisper models for Portuguese 🇧🇷🇵🇹 and have achieved major results. I'm applying for this grant to share these results with the community.
Our best model yet, which uses the Whisper medium architecture, achieved a WER of 6.579 on Common Voice 11 (Portuguese). This is a great improvement over the original Whisper medium (8.100) and Whisper large-v1 (7.100) models. Its accuracy almost reaches that of Whisper large-v2 (6.300) while maintaining the benefits of being a smaller model.
This space currently hosts jlondonobo/whisper-medium-pt. I plan to upgrade it to jlondonobo/whisper-large-v2-pt once it finishes training. I expect the later model to achieve State of the Art accuracy.
Hi
@jlondonobo
- This is fantastic! Your model looks great, thank you for putting in time and effort into fine-tuning the model. We've decided to assign you a T4 instance for the demo. This grant will run till 21st December 2022. After which we'll continue the grant if there's enough traction on the space.
Can't wait to see what you'll do in your next experiments and how you'd bring the WER down further. Thanks again! 🔥
Don't forget to tweet/ LinkedIn your space, I'm sure the community would love to try it. Tag @huggingface and @lambdaapi when you do! 🤗