jonatasgrosman/wav2vec2-large-xlsr-53-russian
🧠 AI Modeljonatasgrosman
Fine-tuned wav2vec2-large for Russian speech recognition with 2.5M downloads.
jonatasgrosman/wav2vec2-large-xlsr-53-russian is an open-source automatic speech recognition model fine-tuned specifically for the Russian language. It uses the wav2vec2-large-xlsr-53 architecture, which leverages self-supervised learning on unlabeled speech data and then fine-tunes for ASR. The model was fine-tuned on the Mozilla Common Voice 6.0 Russian dataset. It supports inference with PyTorch, JAX, and TensorFlow. Key features include a 300M parameter model, tokenization via Wav2Vec2CTCTokenizer, and competitive performance on the HF ASR leaderboard for Russian. With over 2.5 million downloads, it is a widely used resource for Russian speech recognition applications.
💡Highlights
- ├─300M parameter wav2vec2 model
- ├─Fine-tuned on Common Voice 6.0 Russian
- └─2.5M+ downloads from HuggingFace
🎯For
- ├─Speech recognition researchers
- ├─NLP developers
- └─Russian language applications