|
version: "1.34" |
|
|
|
lr: 1e-4 |
|
log_step: 1 |
|
split: 0 |
|
batch_size: 8 |
|
sr: 16000 |
|
|
|
datasets: |
|
- jamendo |
|
- emomusic |
|
- pmemo |
|
- deam |
|
|
|
model: |
|
encoder: "MERT" |
|
layers: |
|
- 5 |
|
- 6 |
|
classifier: "linear-mt-attn-ck" |
|
|
|
|
|
|
|
|
|
kd: True |
|
kd_weight: 0.8 |
|
kd_temperature: 1 |
|
lr: 1e-4 |
|
|
|
dataset: |
|
jamendo: |
|
root: './dataset/jamendo' |
|
subset: 'moodtheme' |
|
batch_size: 8 |
|
output_size : 56 |
|
split: 0 |
|
segment_type: "all" |
|
num_workers: 4 |
|
deam: |
|
root: './dataset/deam' |
|
batch_size: 8 |
|
output_size : 2 |
|
segment_type: "all" |
|
num_workers: 4 |
|
pmemo: |
|
root: './dataset/pmemo' |
|
batch_size: 8 |
|
output_size : 2 |
|
segment_type: "all" |
|
num_workers: 4 |
|
emomusic: |
|
root: './dataset/emomusic' |
|
batch_size: 8 |
|
output_size : 2 |
|
segment_type: "all" |
|
num_workers: 4 |
|
|
|
|
|
genre_class_size: 87 |
|
mood_class_size: 56 |
|
instr_class_size: 40 |
|
dac_latents_size: 72 |
|
dac_rvq_size: 9 |
|
|
|
|
|
checkpoint_pmemo: "tb_logs/best/P.ckpt" |
|
checkpoint_deam: "tb_logs/best/D.ckpt" |
|
checkpoint_emomusic: "tb_logs/best/E.ckpt" |
|
checkpoint_jamendo: "tb_logs/best/J.ckpt" |
|
|
|
|
