Commit
•
d959f41
1
Parent(s):
0dc8264
added CLAP model
Browse files- CLAP/CLAP_weights_2022.pth +3 -0
- CLAP/config.yml +26 -0
CLAP/CLAP_weights_2022.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5b917c47fec10547b65cf1d9096e194896fa1c2d6180536c8524289721a4c5b8
|
| 3 |
+
size 2333972488
|
CLAP/config.yml
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# TEXT ENCODER CONFIG
|
| 2 |
+
text_model: 'bert-base-uncased'
|
| 3 |
+
text_len: 100
|
| 4 |
+
transformer_embed_dim: 768
|
| 5 |
+
freeze_text_encoder_weights: True
|
| 6 |
+
|
| 7 |
+
# AUDIO ENCODER CONFIG
|
| 8 |
+
audioenc_name: 'Cnn14'
|
| 9 |
+
out_emb: 2048
|
| 10 |
+
sampling_rate: 44100
|
| 11 |
+
duration: 5
|
| 12 |
+
fmin: 50
|
| 13 |
+
fmax: 14000
|
| 14 |
+
n_fft: 1028
|
| 15 |
+
hop_size: 320
|
| 16 |
+
mel_bins: 64
|
| 17 |
+
window_size: 1024
|
| 18 |
+
|
| 19 |
+
# PROJECTION SPACE CONFIG
|
| 20 |
+
d_proj: 1024
|
| 21 |
+
temperature: 0.003
|
| 22 |
+
|
| 23 |
+
# TRAINING AND EVALUATION CONFIG
|
| 24 |
+
num_classes: 527
|
| 25 |
+
batch_size: 1024
|
| 26 |
+
demo: False
|