Use the NGC CLI to download:
Copied!
Model
Self-Supervised Learning (SSL) checkpoints for Conformer Large model. These are similar to w2v-Conformer model and can be fine-tuned for Automatic Speech Recognition (ASR).
Use the NGC CLI to download:
Copied!
2 Versions
05/26/2022 11:26 PM UTC616.77 MBCopied!
05/26/2022 11:26 PM UTC616.77 MB
Copied!
Accuracy Librispeech-100h
| Key | Value |
|---|---|
| Librispeech test-other | 6.17 % |
| Librispeech dev-other | 6.08 % |
| Librispeech dev-clean | 3.04 % |
| Librispeech test-clean | 3.17 % |
Accuracy Librispeech-960h
| Key | Value |
|---|---|
| Librispeech Test-other | 4.20 % |
| Librispeech Dev-other | 4.33 % |
| Librispeech Dev-clean | 1.91 % |
| Librispeech Test-clean | 1.96 % |
Model
| Key | Value |
|---|---|
| Encoder Dimension | 512 |
| Number of Encoder Layers | 17 |
| Dataset | Librilight |
| ARCHITECTURE | Conformer |
| INPUTS | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| Number of Weights | 120M |
1.10.0Selected05/26/2022 8:56 PM UTC439.99 MBCopied!
1.10.0Selected
05/26/2022 8:56 PM UTC439.99 MB
Copied!
Accuracy Librispeech-100h
| Key | Value |
|---|---|
| LibrispeechTestOther | 6.39 % |
| LibrispeechDevClean | 3.25 % |
| LibrispeechDevOther | 6.64 % |
| LibrispeechTestClean | 3.19 % |
Accuracy Librispeech-960h
| Key | Value |
|---|---|
| LibrispeechTestOther | 4.50 % |
| LibrispeechDevClean | 1.92 % |
| LibrispeechDevOther | 4.38 % |
| LibrispeechTestClean | 2.02 % |
Model
| Key | Value |
|---|---|
| Encoder Dimension | 512 |
| Number of Encoder Layers | 17 |
| Dataset | Librilight |
| ARCHITECTURE | Conformer |
| INPUTS | 16000 KHZ MONO-CHANNEL AUDIO (WAV FILES) |
| Number of Weights | 120M |