Commit
•
0450d3d
1
Parent(s):
d8df66e
Upload 2 files
Browse files- ttmr_pp.pth +3 -0
- ttmr_pp.yaml +34 -0
ttmr_pp.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7c4b6312898e6a3da97a48a65cb57b419014fa5c374fe67b2cd983d3172c5c2
|
| 3 |
+
size 1944005416
|
ttmr_pp.yaml
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
framework: pretrain
|
| 2 |
+
data_dir: ../../dataset
|
| 3 |
+
train_data: all
|
| 4 |
+
caption_type: meta_tag_caption_sim
|
| 5 |
+
workers: 4
|
| 6 |
+
total_steps: 32768
|
| 7 |
+
start_steps: 0
|
| 8 |
+
batch_size: 768
|
| 9 |
+
world_size: 1
|
| 10 |
+
lr: 5.0e-05
|
| 11 |
+
min_lr: 1.0e-09
|
| 12 |
+
seed: null
|
| 13 |
+
print_freq: 10
|
| 14 |
+
cos: true
|
| 15 |
+
n_fft: 1024
|
| 16 |
+
hop_size: 0.01
|
| 17 |
+
sr: 22050
|
| 18 |
+
duration: 10
|
| 19 |
+
max_length: 128
|
| 20 |
+
audio_loader: ffmpeg
|
| 21 |
+
audio_arch: resnet
|
| 22 |
+
text_arch: roberta-base
|
| 23 |
+
n_heads: 8
|
| 24 |
+
width: 64
|
| 25 |
+
n_mels: 128
|
| 26 |
+
audio_dim: 768
|
| 27 |
+
text_dim: 768
|
| 28 |
+
mlp_dim: 128
|
| 29 |
+
temperature: 0.1
|
| 30 |
+
tid: base
|
| 31 |
+
gpu: 0
|
| 32 |
+
epochs: 19
|
| 33 |
+
start_epoch: 0
|
| 34 |
+
warmup_steps: 5000
|