Instructions to use starkdmi/FRCRN_SE_16K_MLX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use starkdmi/FRCRN_SE_16K_MLX with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] huggingface-cli download --local-dir FRCRN_SE_16K_MLX starkdmi/FRCRN_SE_16K_MLX
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Upload FRCRN_SE_16K.yaml with huggingface_hub
Browse files- FRCRN_SE_16K.yaml +20 -0
FRCRN_SE_16K.yaml
ADDED
|
@@ -0,0 +1,20 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/bin/bash
|
| 2 |
+
mode: 'inference'
|
| 3 |
+
use_cuda: 1 # 1 for True, 0 for False
|
| 4 |
+
num_gpu: 1
|
| 5 |
+
sampling_rate: 16000
|
| 6 |
+
network: "FRCRN_SE_16K" ##network type
|
| 7 |
+
checkpoint_dir: "checkpoints/FRCRN_SE_16K"
|
| 8 |
+
|
| 9 |
+
input_path: "samples/scp/cv_demand_testset_16k.scp" # an input dir or input scp file
|
| 10 |
+
output_dir: "outputs/FRCRN_SE_16K" ## output dir to store processed audio
|
| 11 |
+
|
| 12 |
+
# decode parameters
|
| 13 |
+
one_time_decode_length: 120 #maximum segment length for one-pass decoding (seconds), longer audio will use segmented decoding
|
| 14 |
+
decode_window: 1 #one-pass decoding length
|
| 15 |
+
#
|
| 16 |
+
# FFT parameters
|
| 17 |
+
win_type: 'hanning'
|
| 18 |
+
win_len: 640
|
| 19 |
+
win_inc: 320
|
| 20 |
+
fft_len: 640
|