Audio-to-Audio
MLX
Safetensors
speech-enhancement
speech-denoising
gtcrn
apple-silicon
onnx-conversion
Instructions to use agentable/gtcrn-simple-mlx with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use agentable/gtcrn-simple-mlx with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir gtcrn-simple-mlx agentable/gtcrn-simple-mlx
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
add GTCRN protocol metadata
Browse files- README.md +3 -0
- metadata.json +53 -0
README.md
CHANGED
|
@@ -25,6 +25,9 @@ required by MLX convolutions.
|
|
| 25 |
`d503ed83a35b91f66af97a6a1b07a76425988db7d9744556d0565c336a7e6e3b`.
|
| 26 |
- `conversion.json` - deterministic conversion record binding the source ONNX,
|
| 27 |
audit, converter revision, weight SHA-256, byte count, and 179 tensors.
|
|
|
|
|
|
|
|
|
|
| 28 |
- `LICENSE` - the upstream GTCRN MIT license.
|
| 29 |
|
| 30 |
## Provenance
|
|
|
|
| 25 |
`d503ed83a35b91f66af97a6a1b07a76425988db7d9744556d0565c336a7e6e3b`.
|
| 26 |
- `conversion.json` - deterministic conversion record binding the source ONNX,
|
| 27 |
audit, converter revision, weight SHA-256, byte count, and 179 tensors.
|
| 28 |
+
- `metadata.json` - GTCRN Simple protocol metadata shared with the pinned ONNX
|
| 29 |
+
asset. Its SHA-256 is
|
| 30 |
+
`070f320defc15d4e1781b0c5b9225d9f91031ed5f14f675efa497f244e640615`.
|
| 31 |
- `LICENSE` - the upstream GTCRN MIT license.
|
| 32 |
|
| 33 |
## Provenance
|
metadata.json
ADDED
|
@@ -0,0 +1,53 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"protocol": "gtcrn-simple-v1",
|
| 3 |
+
"model_type": "gtcrn",
|
| 4 |
+
"sample_rate": 16000,
|
| 5 |
+
"resample": {
|
| 6 |
+
"kind": "sherpa-linear-resample",
|
| 7 |
+
"lowpass_cutoff_ratio": 0.99,
|
| 8 |
+
"lowpass_filter_width": 6,
|
| 9 |
+
"flush": true
|
| 10 |
+
},
|
| 11 |
+
"stft": {
|
| 12 |
+
"n_fft": 512,
|
| 13 |
+
"hop_length": 256,
|
| 14 |
+
"window_length": 512,
|
| 15 |
+
"window_type": "hann_sqrt",
|
| 16 |
+
"bins": 257
|
| 17 |
+
},
|
| 18 |
+
"input": {
|
| 19 |
+
"name": "mix",
|
| 20 |
+
"dtype": "float32",
|
| 21 |
+
"shape": [1, 257, 1, 2],
|
| 22 |
+
"complex_layout": "real_imag"
|
| 23 |
+
},
|
| 24 |
+
"caches": [
|
| 25 |
+
{
|
| 26 |
+
"input_name": "conv_cache",
|
| 27 |
+
"output_name": "conv_cache_out",
|
| 28 |
+
"dtype": "float32",
|
| 29 |
+
"shape": [2, 1, 16, 16, 33]
|
| 30 |
+
},
|
| 31 |
+
{
|
| 32 |
+
"input_name": "tra_cache",
|
| 33 |
+
"output_name": "tra_cache_out",
|
| 34 |
+
"dtype": "float32",
|
| 35 |
+
"shape": [2, 3, 1, 1, 16]
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"input_name": "inter_cache",
|
| 39 |
+
"output_name": "inter_cache_out",
|
| 40 |
+
"dtype": "float32",
|
| 41 |
+
"shape": [2, 1, 33, 16]
|
| 42 |
+
}
|
| 43 |
+
],
|
| 44 |
+
"output": {
|
| 45 |
+
"name": "enh",
|
| 46 |
+
"dtype": "float32",
|
| 47 |
+
"shape": [1, 257, 1, 2],
|
| 48 |
+
"complex_layout": "real_imag"
|
| 49 |
+
},
|
| 50 |
+
"limits": {
|
| 51 |
+
"max_output_samples": 4800000
|
| 52 |
+
}
|
| 53 |
+
}
|