Guillermo Cámbara

icon indicating a website link https://www.linkedin.com/in/guillermo-cámbara-ruiz-43312a68/ icon indicating an email [email protected]

icon company Universitat Pompeu Fabra icon location Barcelona, Spain icon location 🎙️ PhD Candidate in Self-Supervised Learning + Speech Recognition @ Universitat Pompeu Fabra & Telefónica Research

Results 1 comments of Guillermo Cámbara

streaming_convnets model. the WER does not decrease during training (it remains stable at 100% for dev and train)

I've had the same issue with the streaming_convnets word-piece model applied to LibriSpeech 100h, which I've managed to solve by tweaking the SpecAugment parameters in the architecture file. First, I've...