camenduru commited on
Commit
d159f33
1 Parent(s): 0096e34

thanks to ymzhang319 ❤

Browse files
README.md ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
semantic/semantic_adapter.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87db0038a5581334c0785d67a41dc7290aee4c3ad7f3564b96423366bbc1935c
3
+ size 89289954
temporal_adapter.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54575474c69a0c949bddbdc85fc9d5c2e560eb9cf0b3bed749da4538f62262f5
3
+ size 4335749362
timestamp_detector.pth.tar ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4fa4fba46c38e16a3a96825a55c3690b4a72ffbb337f6eed11dc1eadf79a78cf
3
+ size 376654135
vocoder/config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "resblock": "1",
3
+ "num_gpus": 0,
4
+ "batch_size": 16,
5
+ "learning_rate": 0.0002,
6
+ "adam_b1": 0.8,
7
+ "adam_b2": 0.99,
8
+ "lr_decay": 0.999,
9
+ "seed": 1234,
10
+
11
+ "upsample_rates": [5,4,4,2],
12
+ "upsample_kernel_sizes": [11,8,8,4],
13
+ "upsample_initial_channel": 512,
14
+ "resblock_kernel_sizes": [3,7,11],
15
+ "resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
16
+
17
+ "segment_size": 5120,
18
+ "num_mels": 256,
19
+ "num_freq": 2049,
20
+ "n_fft": 2048,
21
+ "hop_size": 160,
22
+ "win_size": 1024,
23
+
24
+ "sampling_rate": 16000,
25
+
26
+ "fmin": 0,
27
+ "fmax": null,
28
+ "fmax_for_loss": null,
29
+
30
+ "num_workers": 4,
31
+
32
+ "dist_config": {
33
+ "dist_backend": "nccl",
34
+ "dist_url": "tcp://localhost:54321",
35
+ "world_size": 1
36
+ }
37
+ }
vocoder/vocoder.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d71bb8ec87c21902175cb0f918e739789ec085663ed0f8ebd2f20adba4d5b5af
3
+ size 54808625