Initial commit
Browse files- README.md +86 -0
- pytorch_model.bin +3 -0
README.md
ADDED
|
@@ -0,0 +1,86 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
tags:
|
| 3 |
+
- asteroid
|
| 4 |
+
- audio
|
| 5 |
+
- ConvTasNet
|
| 6 |
+
datasets:
|
| 7 |
+
- Libri1Mix
|
| 8 |
+
- enh_single
|
| 9 |
+
license: CC BY-NC 4.0
|
| 10 |
+
inference: false
|
| 11 |
+
---
|
| 12 |
+
|
| 13 |
+
## Asteroid model `JorisCos/DPTNet_Libri1Mix_enhsignle_16k`
|
| 14 |
+
|
| 15 |
+
Description:
|
| 16 |
+
|
| 17 |
+
This model was trained by Joris Cosentino using the librimix recipe in [Asteroid](https://github.com/asteroid-team/asteroid).
|
| 18 |
+
It was trained on the `enh_single` task of the Libri1Mix dataset.
|
| 19 |
+
|
| 20 |
+
Training config:
|
| 21 |
+
|
| 22 |
+
```yml
|
| 23 |
+
data:
|
| 24 |
+
n_src: 1
|
| 25 |
+
sample_rate: 16000
|
| 26 |
+
segment: 3
|
| 27 |
+
task: enh_single
|
| 28 |
+
train_dir: data/wav16k/min/train-360
|
| 29 |
+
valid_dir: data/wav16k/min/dev
|
| 30 |
+
filterbank:
|
| 31 |
+
kernel_size: 16
|
| 32 |
+
n_filters: 64
|
| 33 |
+
stride: 8
|
| 34 |
+
masknet:
|
| 35 |
+
bidirectional: true
|
| 36 |
+
chunk_size: 100
|
| 37 |
+
dropout: 0
|
| 38 |
+
ff_activation: relu
|
| 39 |
+
ff_hid: 256
|
| 40 |
+
hop_size: 50
|
| 41 |
+
in_chan: 64
|
| 42 |
+
mask_act: sigmoid
|
| 43 |
+
n_repeats: 2
|
| 44 |
+
n_src: 1
|
| 45 |
+
norm_type: gLN
|
| 46 |
+
out_chan: 64
|
| 47 |
+
optim:
|
| 48 |
+
lr: 0.001
|
| 49 |
+
optimizer: adam
|
| 50 |
+
weight_decay: 1.0e-05
|
| 51 |
+
scheduler:
|
| 52 |
+
d_model: 64
|
| 53 |
+
steps_per_epoch: 10000
|
| 54 |
+
training:
|
| 55 |
+
batch_size: 4
|
| 56 |
+
early_stop: true
|
| 57 |
+
epochs: 200
|
| 58 |
+
gradient_clipping: 5
|
| 59 |
+
half_lr: true
|
| 60 |
+
num_workers: 4
|
| 61 |
+
```
|
| 62 |
+
|
| 63 |
+
|
| 64 |
+
Results:
|
| 65 |
+
|
| 66 |
+
On Libri1Mix min test set :
|
| 67 |
+
```yml
|
| 68 |
+
si_sdr: 14.829670037349064,
|
| 69 |
+
si_sdr_imp: 11.379888731489366,
|
| 70 |
+
sdr: 15.395712644737149,
|
| 71 |
+
sdr_imp: 11.893049845524112,
|
| 72 |
+
sir: Infinity,
|
| 73 |
+
sir_imp: NaN,
|
| 74 |
+
sar: 15.395712644737149,
|
| 75 |
+
sar_imp: 11.893049845524112,
|
| 76 |
+
stoi: 0.9301948391058859,
|
| 77 |
+
stoi_imp: 0.13427501556534832
|
| 78 |
+
```
|
| 79 |
+
|
| 80 |
+
|
| 81 |
+
License notice:
|
| 82 |
+
|
| 83 |
+
This work "DPTNet_Libri1Mix_enhsignle_16k" is a derivative of [LibriSpeech ASR corpus](http://www.openslr.org/12) by Vassil Panayotov,
|
| 84 |
+
used under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/); of The WSJ0 Hipster Ambient Mixtures
|
| 85 |
+
dataset by [Whisper.ai](http://wham.whisper.ai/), used under [CC BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) (Research only).
|
| 86 |
+
"DPTNet_Libri1Mix_enhsignle_16k" is licensed under [Attribution-ShareAlike 3.0 Unported](https://creativecommons.org/licenses/by-sa/3.0/) by Joris Cosentino
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f4f38dc0be2bcb479364b4b49fdc0c92d77fc3f1aa6049090cd3ea0db95019f
|
| 3 |
+
size 11437018
|