par facebook
Open source · 42k downloads · 35 likes
Hubert Large LL60k est un modèle de traitement automatique de la parole conçu pour apprendre des représentations audio de manière auto-supervisée. Il excelle dans la reconnaissance vocale, la génération et la compression de la parole, grâce à une approche innovante qui combine apprentissage acoustique et modélisation linguistique. Pour être utilisé en reconnaissance vocale, il nécessite un fine-tuning avec des données textuelles labellisées et un tokenizer adapté, car il n’intègre pas de tokenizer par défaut. Ce modèle se distingue par sa capacité à gérer des unités sonores variables et à s’adapter à différentes quantités de données d’entraînement, offrant des performances de pointe sur des benchmarks comme Librispeech et Libri-light. Son architecture, basée sur des itérations de clustering non supervisé, le rend particulièrement robuste pour des applications nécessitant une compréhension fine de la parole.
The large model pretrained on 16kHz sampled speech audio. When using the model make sure that your speech input is also sampled at 16Khz.
Note: This model does not have a tokenizer as it was pretrained on audio alone. In order to use this model speech recognition, a tokenizer should be created and the model should be fine-tuned on labeled text data. Check out this blog for more in-detail explanation of how to fine-tune the model.
The model was pretrained on Libri-Light.
Authors: Wei-Ning Hsu, Benjamin Bolte, Yao-Hung Hubert Tsai, Kushal Lakhotia, Ruslan Salakhutdinov, Abdelrahman Mohamed
Abstract Self-supervised approaches for speech representation learning are challenged by three unique problems: (1) there are multiple sound units in each input utterance, (2) there is no lexicon of input sound units during the pre-training phase, and (3) sound units have variable lengths with no explicit segmentation. To deal with these three problems, we propose the Hidden-Unit BERT (HuBERT) approach for self-supervised speech representation learning, which utilizes an offline clustering step to provide aligned target labels for a BERT-like prediction loss. A key ingredient of our approach is applying the prediction loss over the masked regions only, which forces the model to learn a combined acoustic and language model over the continuous inputs. HuBERT relies primarily on the consistency of the unsupervised clustering step rather than the intrinsic quality of the assigned cluster labels. Starting with a simple k-means teacher of 100 clusters, and using two iterations of clustering, the HuBERT model either matches or improves upon the state-of-the-art wav2vec 2.0 performance on the Librispeech (960h) and Libri-light (60,000h) benchmarks with 10min, 1h, 10h, 100h, and 960h fine-tuning subsets. Using a 1B parameter model, HuBERT shows up to 19% and 13% relative WER reduction on the more challenging dev-other and test-other evaluation subsets.
The original model can be found under https://github.com/pytorch/fairseq/tree/master/examples/hubert .
See this blog for more information on how to fine-tune the model. Note that the class Wav2Vec2ForCTC has to be replaced by HubertForCTC.
