|
|
--- |
|
|
library_name: transformers |
|
|
pipeline_tag: automatic-speech-recognition |
|
|
language: en |
|
|
license: mit |
|
|
--- |
|
|
|
|
|
# Whisper MLX Model |
|
|
|
|
|
This repository contains a CoreML and MLX-optimized Whisper model for efficient speech recognition on Apple devices. |
|
|
|
|
|
## Model Components |
|
|
|
|
|
- CoreML encoder for efficient inferencing on Apple Neural Engine |
|
|
- MLX decoder for fast processing with Apple Silicon optimizations |
|
|
- HuggingFace Whisper processor and model template |
|
|
|
|
|
## Usage |
|
|
|
|
|
```python |
|
|
from asr_streaming import StreamingConfig, StreamingBackend |
|
|
|
|
|
# Create config with HuggingFace repository |
|
|
config = StreamingConfig( |
|
|
use_huggingface=True, |
|
|
huggingface_repo="TheStageAI/whisper-medium" |
|
|
) |
|
|
|
|
|
# Initialize the backend |
|
|
backend = StreamingBackend(config) |
|
|
|
|
|
# Now use the backend as normal |
|
|
``` |
|
|
|
