Instructions to use videodanchik/icefall-asr-tedlium3-conformer-ctc2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- K2
How to use videodanchik/icefall-asr-tedlium3-conformer-ctc2 with K2:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Tedlium3
This is an ASR ctc attention encoder-decoder recipe for the Tedlium3 corpus. The encoder is the re-worked conformer model. Language models are downloaded from https://kaldi-asr.org/models/m5
Performance Record
conformer_ctc2
| Decodeing method | dev WER | test WER |
|---|---|---|
| ctc-decoding | 6.45 | 5.96 |
| 1best | 5.92 | 5.51 |
| whole-lattice-rescoring | 5.96 | 5.47 |
| attention-decoder | 5.60 | 5.33 |
See the recipe for details.