Views
No views yet
JorisCos/ConvTasNet_Libri2Mix_sepnoisy_8k.1# Konfigurasi yang digunakan selama fine-tuning
2data:
3 root: "data/processed/"
4 sample_rate: 8000
5 segment_seconds: 4
6 num_workers: 4
7
8training:
9 project_name: "itera-speech-separation-ft"
10 model_name: "ConvTasNet-ITERA-FT" # Nama yang digunakan selama training
11 epochs: 50
12 batch_size: 8
13 learning_rate: 0.0005
14 gradient_clip_val: 0.5
15 precision: "16-mixed"
16 early_stopping_patience: 5
17
18model:
19 freeze_encoder_decoder: false
20
21remix:
22 dynamic: true
23 snr_low: 0.0
24 snr_high: 10.01si_sdr:
2 baseline_score: -30.2842
3 fine_tuned_score: -24.9016
4 improvement: +5.3826JorisCos/ConvTasNet_Libri2Mix_sepnoisy_8k. The original work is a derivative of:
- LibriSpeech ASR corpus by Vassil Panayotov, used under CC BY 4.0;
- The WSJ0 Hipster Ambient Mixtures dataset by Whisper.ai, used under CC BY-NC 4.0.
The original work is licensed under Attribution-ShareAlike 3.0 Unported by Joris Cosentino.