metadata
tags:
- asteroid
- audio
- DPTNet
- audio-to-audio
datasets:
- Libri1Mix
- enh_single
license: cc-by-sa-4.0
Asteroid model cankeles/DPTNet_WHAMR_enhsignle_16k
Description:
This model was trained by M. Can Keleş using the librimix recipe in Asteroid.
It was trained on the enh_single task of the Libri1Mix dataset.
Training config:
data:
mode: min
nondefault_nsrc: null
sample_rate: 16000
segment: 2.0
task: enh_single
train_dir: wav16k/min/tr/
valid_dir: wav16k/min/cv/
filterbank:
kernel_size: 16
n_filters: 64
stride: 8
main_args:
exp_dir: exp/tmp
help: null
masknet:
bidirectional: true
chunk_size: 100
dropout: 0
ff_activation: relu
ff_hid: 256
hop_size: 50
in_chan: 64
mask_act: sigmoid
n_repeats: 2
n_src: 1
norm_type: gLN
out_chan: 64
optim:
lr: 0.001
optimizer: adam
weight_decay: 1.0e-05
positional arguments: {}
scheduler:
d_model: 64
steps_per_epoch: 10000
training:
batch_size: 4
early_stop: true
epochs: 60
gradient_clipping: 5
half_lr: true
num_workers: 4
Results:
On custom min test set :
'sar': 12.853384266251018,
'sar_imp': 8.950332361953906,
'sdr': 12.853384266251018,
'sdr_imp': 8.950332361953906,
'si_sdr': 12.247012621312548,
'si_sdr_imp': 8.429646186633407,
'sir': inf,
'sir_imp': nan,
'stoi': 0.9022338865380519,
'stoi_imp': 0.09735707619500522