update readme, add links, and metrics
#1
by mlobdell-synaptics - opened
README.md
CHANGED
|
@@ -19,21 +19,41 @@ Moonshine is a high-efficiency automatic speech recognition (ASR) model designed
|
|
| 19 |
|
| 20 |
Useful Sensors developed Moonshine and released the English model as open-source. There are 2 models of different sizes and capabilities - base and tiny. The tiny version utilizes 27M parameters.
|
| 21 |
|
| 22 |
-
|
| 23 |
-
Moonshine Tiny has been optimized for the Synaptics Astra™ **SL2610-Series processors** with Torq NPU.
|
| 24 |
-
|
| 25 |
## Model Features
|
| 26 |
- **Model Type:** Automatic Speech Recognition
|
| 27 |
- **Input:** Raw waveform (1D array of floats) 16kHz mono audio up to 30 seconds
|
| 28 |
- **Output:** Sequence of token IDs (integers)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 29 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
|
| 31 |
## Deployment
|
| 32 |
|
| 33 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 34 |
|
| 35 |
-
|
|
|
|
|
|
|
| 36 |
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
## License
|
| 39 |
|
|
|
|
| 19 |
|
| 20 |
Useful Sensors developed Moonshine and released the English model as open-source. There are 2 models of different sizes and capabilities - base and tiny. The tiny version utilizes 27M parameters.
|
| 21 |
|
|
|
|
|
|
|
|
|
|
| 22 |
## Model Features
|
| 23 |
- **Model Type:** Automatic Speech Recognition
|
| 24 |
- **Input:** Raw waveform (1D array of floats) 16kHz mono audio up to 30 seconds
|
| 25 |
- **Output:** Sequence of token IDs (integers)
|
| 26 |
+
- **Quantization**: None
|
| 27 |
+
|
| 28 |
+
## Recommended Platforms
|
| 29 |
+
|
| 30 |
+
- Synaptics Astra™ SL2600-Series with Torq
|
| 31 |
+
- Synaptics Astra™ SL1600-Series with Synap
|
| 32 |
|
| 33 |
+
## Metrics
|
| 34 |
+
|
| 35 |
+
| Platform | Model / Stage | Environment | Inference Time | Infer / s |
|
| 36 |
+
| --------- | --------- | --------- | --------- | --------- |
|
| 37 |
+
| SL2610 | Moonshine Tiny Decoder | Torq v2.0.0 | 15.6 | 64.1 |
|
| 38 |
+
| SL2610 | Moonshine Tiny Encoder (5 sec audio) | Torq v2.0.0 | 156.1 | 6.4
|
| 39 |
|
| 40 |
## Deployment
|
| 41 |
|
| 42 |
+
### Compiled Models
|
| 43 |
+
Torq compiled model files are provided in this repository. To recompile the models, see the [Torq Documentation](https://synaptics-torq.github.io/torq-compiler/v/latest/).
|
| 44 |
+
|
| 45 |
+
### Source Models
|
| 46 |
+
The source model files are available at [Synaptics/Moonshine](https://huggingface.co/Synaptics/Moonshine).
|
| 47 |
+
|
| 48 |
+
### Usage Tutorials / Example Apps
|
| 49 |
|
| 50 |
+
Synaptics AI Developer Zone Tutorials
|
| 51 |
+
- [SL2600-Series Speech-To-Text Tutorial](https://developer.synaptics.com/docs/sl/sl2600/introduction)
|
| 52 |
+
- [SL1600-Series Speech-To-Text Tutorial](https://developer.synaptics.com/docs/sl/sl1600/getting-started/speech-to-text#useful-sensorss-moonshine)
|
| 53 |
|
| 54 |
+
Example App GitHub Repositories
|
| 55 |
+
- [Torq Examples](https://github.com/synaptics-torq/torq-examples): Basic model usage examples for Torq
|
| 56 |
+
- [SL2610 Examples](https://github.com/synaptics-astra-demos/sl2610-examples): Interactive examples for SL2610-Series
|
| 57 |
|
| 58 |
## License
|
| 59 |
|