TF-GridNet (code, models, paper)
Browse files- .gitattributes +2 -0
- A time-frequency domain target speaker extraction network with adaptive speaker embedding fusion.pdf +3 -0
- code/X-TF-GridNet.zip +3 -0
- code/tf_gridnet_demo.zip +3 -0
- models/pTFGridNet_WHAMR/checkpoint/best.pt.tar +3 -0
- models/pTFGridNet_WHAMR/config.json +72 -0
- models/pTFGridNet_WHAMR/loss_mat.mat +3 -0
- models/pTFGridNet_WHAMR/net_config.json +11 -0
- models/pTFGridNet_WHAMR/source.txt +1 -0
- models/pTFGridNet_WHAMR/trainer.log +0 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
A[[:space:]]time-frequency[[:space:]]domain[[:space:]]target[[:space:]]speaker[[:space:]]extraction[[:space:]]network[[:space:]]with[[:space:]]adaptive[[:space:]]speaker[[:space:]]embedding[[:space:]]fusion.pdf filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
models/pTFGridNet_WHAMR/loss_mat.mat filter=lfs diff=lfs merge=lfs -text
|
A time-frequency domain target speaker extraction network with adaptive speaker embedding fusion.pdf
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48a813cda0707362decac42a7154bcaced18ff81b2a57b6483edc02ebd781699
|
| 3 |
+
size 21682680
|
code/X-TF-GridNet.zip
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc912b92b3c8cf69c7598f199f069a760e0e3a7e393fa000d03607e260d4940b
|
| 3 |
+
size 130299949
|
code/tf_gridnet_demo.zip
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2803f7bb8949916dab8422ea905adbf15db41637d10572e0867be8c082a5419d
|
| 3 |
+
size 1141575
|
models/pTFGridNet_WHAMR/checkpoint/best.pt.tar
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74bab3a83424c3f381ef1b7898cdcb038469baf4abb79dc28dbdffcd794b5f72
|
| 3 |
+
size 93985267
|
models/pTFGridNet_WHAMR/config.json
ADDED
|
@@ -0,0 +1,72 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"path": {
|
| 3 |
+
"num_spks": 101,
|
| 4 |
+
"resume_filename": "",
|
| 5 |
+
"train": {
|
| 6 |
+
"scp_dir": "data/whamr_mr/tr/",
|
| 7 |
+
"mix_scp": "mix.scp",
|
| 8 |
+
"ref_scp": "ref.scp",
|
| 9 |
+
"aux_scp": "aux.scp",
|
| 10 |
+
"spk_list": "data/wsj0_2mix_extr_tr.spk",
|
| 11 |
+
"ref_dur": "data/uniq_target_ref_dur.txt"
|
| 12 |
+
},
|
| 13 |
+
"val": {
|
| 14 |
+
"scp_dir": "data/whamr_mr/cv/",
|
| 15 |
+
"mix_scp": "mix.scp",
|
| 16 |
+
"ref_scp": "ref.scp",
|
| 17 |
+
"aux_scp": "aux.scp",
|
| 18 |
+
"spk_list": "data/wsj0_2mix_extr_tr.spk",
|
| 19 |
+
"ref_dur": "data/uniq_target_ref_dur.txt"
|
| 20 |
+
}
|
| 21 |
+
},
|
| 22 |
+
"gpu": {
|
| 23 |
+
"gpu_ids": [
|
| 24 |
+
2
|
| 25 |
+
]
|
| 26 |
+
},
|
| 27 |
+
"signal": {
|
| 28 |
+
"sr": 8000,
|
| 29 |
+
"chunk_length": 4,
|
| 30 |
+
"win_size": 0.032,
|
| 31 |
+
"win_shift": 0.008,
|
| 32 |
+
"fft_num": 256,
|
| 33 |
+
"beta": 1.0
|
| 34 |
+
},
|
| 35 |
+
"loss_function": {
|
| 36 |
+
"zero_mean": true,
|
| 37 |
+
"scale_label": true
|
| 38 |
+
},
|
| 39 |
+
"optimizer": {
|
| 40 |
+
"name": "adam",
|
| 41 |
+
"lr": 0.001,
|
| 42 |
+
"beta1": 0.9,
|
| 43 |
+
"beta2": 0.999,
|
| 44 |
+
"l2": 1e-07,
|
| 45 |
+
"min_lr": 1e-08,
|
| 46 |
+
"logging_period": 1000,
|
| 47 |
+
"factor": 0.5,
|
| 48 |
+
"gradient_norm": 5.0,
|
| 49 |
+
"epochs": 150,
|
| 50 |
+
"halve_freq": 2,
|
| 51 |
+
"early_stop_freq": 6
|
| 52 |
+
},
|
| 53 |
+
"dataloader": {
|
| 54 |
+
"num_workers": 4,
|
| 55 |
+
"pin_memory": true,
|
| 56 |
+
"batch_size": 5
|
| 57 |
+
},
|
| 58 |
+
"net": {
|
| 59 |
+
"n_layers": 3,
|
| 60 |
+
"lstm_hidden_units": 256,
|
| 61 |
+
"attn_n_head": 4,
|
| 62 |
+
"attn_approx_qk_dim": 512,
|
| 63 |
+
"emb_dim": 64,
|
| 64 |
+
"emb_ks": 4,
|
| 65 |
+
"emb_hs": 2,
|
| 66 |
+
"activation": "prelu",
|
| 67 |
+
"eps": 1e-05
|
| 68 |
+
},
|
| 69 |
+
"save": {
|
| 70 |
+
"save_filename": "exp/pTFGridNet_3.5"
|
| 71 |
+
}
|
| 72 |
+
}
|
models/pTFGridNet_WHAMR/loss_mat.mat
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f1556be13a6014f9814c850b0171ea40377045e9633080de37155f22e2ae8cf
|
| 3 |
+
size 177192
|
models/pTFGridNet_WHAMR/net_config.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"n_layers": 3,
|
| 3 |
+
"lstm_hidden_units": 256,
|
| 4 |
+
"attn_n_head": 4,
|
| 5 |
+
"attn_approx_qk_dim": 512,
|
| 6 |
+
"emb_dim": 64,
|
| 7 |
+
"emb_ks": 4,
|
| 8 |
+
"emb_hs": 2,
|
| 9 |
+
"activation": "prelu",
|
| 10 |
+
"eps": 1e-05
|
| 11 |
+
}
|
models/pTFGridNet_WHAMR/source.txt
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
https://github.com/HaoFengyuan/X-TF-GridNet/tree/main/exp/pTFGridNet_WHAMR
|
models/pTFGridNet_WHAMR/trainer.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|