Upload folder using huggingface_hub
Browse files
weights/BEATs_iter3_plus_AS2M.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d43cbfad4d7b56381c061d7a24774f908d4d94c72961f6eb1d9090ff18cd8d34
|
| 3 |
+
size 361499833
|
weights/beats_aasist.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6332dc41bffcde23ca6bf21888f3b8ae899c9c6fa1132986869675fcbb1942aa
|
| 3 |
+
size 13460
|
weights/detection_MHFA_wav2vec2_large/detection_MHFA_wav2vec2_large/avg_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4cb773ea02916eecd32a24529e083bbacdddf8c4dcdd124cb98c9e208af8e17
|
| 3 |
+
size 1274970731
|
weights/detection_MHFA_wav2vec2_large/detection_MHFA_wav2vec2_large/config.yaml
ADDED
|
@@ -0,0 +1,93 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
data_type: shard
|
| 2 |
+
dataloader_args:
|
| 3 |
+
batch_size: 16
|
| 4 |
+
drop_last: true
|
| 5 |
+
num_workers: 6
|
| 6 |
+
pin_memory: false
|
| 7 |
+
prefetch_factor: 8
|
| 8 |
+
dataset_args:
|
| 9 |
+
aug_prob: 0
|
| 10 |
+
cmvn: true
|
| 11 |
+
cmvn_args:
|
| 12 |
+
norm_mean: true
|
| 13 |
+
norm_var: false
|
| 14 |
+
filter: true
|
| 15 |
+
filter_args:
|
| 16 |
+
max_num_frames: 400
|
| 17 |
+
min_num_frames: 50
|
| 18 |
+
frontend: s3prl
|
| 19 |
+
num_frms: 150
|
| 20 |
+
resample_rate: 16000
|
| 21 |
+
s3prl_args:
|
| 22 |
+
download_dir: ./s3prl_hub
|
| 23 |
+
frame_length: 20
|
| 24 |
+
frame_shift: 20
|
| 25 |
+
frozen: false
|
| 26 |
+
layer: -1
|
| 27 |
+
layerwise_feature: true
|
| 28 |
+
multilayer_feature: true
|
| 29 |
+
upstream_args:
|
| 30 |
+
name: wav2vec2_large_960
|
| 31 |
+
sample_num_per_epoch: 0
|
| 32 |
+
shuffle: true
|
| 33 |
+
shuffle_args:
|
| 34 |
+
shuffle_size: 2500
|
| 35 |
+
spec_aug: false
|
| 36 |
+
spec_aug_args:
|
| 37 |
+
max_f: 8
|
| 38 |
+
max_t: 10
|
| 39 |
+
num_f_mask: 1
|
| 40 |
+
num_t_mask: 1
|
| 41 |
+
prob: 0.6
|
| 42 |
+
speed_perturb: false
|
| 43 |
+
enable_amp: false
|
| 44 |
+
exp_dir: exp/MHFA_wav2vec2_large_960-FT-1stage5
|
| 45 |
+
gpus:
|
| 46 |
+
- 1
|
| 47 |
+
log_batch_interval: 100
|
| 48 |
+
loss: CrossEntropyLoss
|
| 49 |
+
loss_args: {}
|
| 50 |
+
margin_scheduler: MarginScheduler
|
| 51 |
+
margin_update:
|
| 52 |
+
epoch_iter: 11397
|
| 53 |
+
final_margin: 0.0
|
| 54 |
+
fix_start_epoch: 1
|
| 55 |
+
increase_start_epoch: 1
|
| 56 |
+
increase_type: exp
|
| 57 |
+
initial_margin: 0.0
|
| 58 |
+
update_margin: true
|
| 59 |
+
model: SSL_BACKEND_MHFA
|
| 60 |
+
model_args:
|
| 61 |
+
compression_dim: 128
|
| 62 |
+
embed_dim: 256
|
| 63 |
+
feat_dim: 1024
|
| 64 |
+
feature_grad_mult: 0.05
|
| 65 |
+
head_nb: 32
|
| 66 |
+
nb_layer: 25
|
| 67 |
+
model_init: null
|
| 68 |
+
num_avg: 2
|
| 69 |
+
num_epochs: 5
|
| 70 |
+
optimizer: AdamW
|
| 71 |
+
optimizer_args:
|
| 72 |
+
lr: 1.0e-05
|
| 73 |
+
weight_decay: 1.0e-08
|
| 74 |
+
projection_args:
|
| 75 |
+
do_lm: false
|
| 76 |
+
easy_margin: false
|
| 77 |
+
embed_dim: 256
|
| 78 |
+
num_class: 2
|
| 79 |
+
project_type: softmax
|
| 80 |
+
scale: 32.0
|
| 81 |
+
save_epoch_interval: 1
|
| 82 |
+
scheduler: ExponentialDecrease
|
| 83 |
+
scheduler_args:
|
| 84 |
+
epoch_iter: 11397
|
| 85 |
+
final_lr: 5.0e-07
|
| 86 |
+
initial_lr: 1.0e-05
|
| 87 |
+
num_epochs: 5
|
| 88 |
+
scale_ratio: 0.25
|
| 89 |
+
warm_from_zero: true
|
| 90 |
+
warm_up_epoch: 1
|
| 91 |
+
seed: 42
|
| 92 |
+
train_data: data/asv5/train/shard.list
|
| 93 |
+
train_label: data/asv5/train/utt2cls
|