added jsons for bigvgan
Browse files
models/bigvgan_24khz_100band.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"resblock": "1",
|
| 3 |
+
"num_gpus": 0,
|
| 4 |
+
"batch_size": 32,
|
| 5 |
+
"learning_rate": 0.0001,
|
| 6 |
+
"adam_b1": 0.8,
|
| 7 |
+
"adam_b2": 0.99,
|
| 8 |
+
"lr_decay": 0.999,
|
| 9 |
+
"seed": 1234,
|
| 10 |
+
|
| 11 |
+
"upsample_rates": [4,4,2,2,2,2],
|
| 12 |
+
"upsample_kernel_sizes": [8,8,4,4,4,4],
|
| 13 |
+
"upsample_initial_channel": 1536,
|
| 14 |
+
"resblock_kernel_sizes": [3,7,11],
|
| 15 |
+
"resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
|
| 16 |
+
|
| 17 |
+
"activation": "snakebeta",
|
| 18 |
+
"snake_logscale": true,
|
| 19 |
+
|
| 20 |
+
"discriminator": "mrd",
|
| 21 |
+
"resolutions": [[1024, 120, 600], [2048, 240, 1200], [512, 50, 240]],
|
| 22 |
+
"mpd_reshapes": [2, 3, 5, 7, 11],
|
| 23 |
+
"use_spectral_norm": false,
|
| 24 |
+
"discriminator_channel_mult": 1,
|
| 25 |
+
|
| 26 |
+
"segment_size": 8192,
|
| 27 |
+
"num_mels": 100,
|
| 28 |
+
"num_freq": 1025,
|
| 29 |
+
"n_fft": 1024,
|
| 30 |
+
"hop_size": 256,
|
| 31 |
+
"win_size": 1024,
|
| 32 |
+
|
| 33 |
+
"sampling_rate": 24000,
|
| 34 |
+
|
| 35 |
+
"fmin": 0,
|
| 36 |
+
"fmax": 12000,
|
| 37 |
+
"fmax_for_loss": null,
|
| 38 |
+
|
| 39 |
+
"num_workers": 4,
|
| 40 |
+
|
| 41 |
+
"dist_config": {
|
| 42 |
+
"dist_backend": "nccl",
|
| 43 |
+
"dist_url": "tcp://localhost:54321",
|
| 44 |
+
"world_size": 1
|
| 45 |
+
}
|
| 46 |
+
}
|
models/bigvgan_base_24khz_100band.json
ADDED
|
@@ -0,0 +1,46 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"resblock": "1",
|
| 3 |
+
"num_gpus": 0,
|
| 4 |
+
"batch_size": 32,
|
| 5 |
+
"learning_rate": 0.0001,
|
| 6 |
+
"adam_b1": 0.8,
|
| 7 |
+
"adam_b2": 0.99,
|
| 8 |
+
"lr_decay": 0.999,
|
| 9 |
+
"seed": 1234,
|
| 10 |
+
|
| 11 |
+
"upsample_rates": [8,8,2,2],
|
| 12 |
+
"upsample_kernel_sizes": [16,16,4,4],
|
| 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 |
+
"activation": "snakebeta",
|
| 18 |
+
"snake_logscale": true,
|
| 19 |
+
|
| 20 |
+
"discriminator": "mrd",
|
| 21 |
+
"resolutions": [[1024, 120, 600], [2048, 240, 1200], [512, 50, 240]],
|
| 22 |
+
"mpd_reshapes": [2, 3, 5, 7, 11],
|
| 23 |
+
"use_spectral_norm": false,
|
| 24 |
+
"discriminator_channel_mult": 1,
|
| 25 |
+
|
| 26 |
+
"segment_size": 8192,
|
| 27 |
+
"num_mels": 100,
|
| 28 |
+
"num_freq": 1025,
|
| 29 |
+
"n_fft": 1024,
|
| 30 |
+
"hop_size": 256,
|
| 31 |
+
"win_size": 1024,
|
| 32 |
+
|
| 33 |
+
"sampling_rate": 24000,
|
| 34 |
+
|
| 35 |
+
"fmin": 0,
|
| 36 |
+
"fmax": 12000,
|
| 37 |
+
"fmax_for_loss": null,
|
| 38 |
+
|
| 39 |
+
"num_workers": 4,
|
| 40 |
+
|
| 41 |
+
"dist_config": {
|
| 42 |
+
"dist_backend": "nccl",
|
| 43 |
+
"dist_url": "tcp://localhost:54321",
|
| 44 |
+
"world_size": 1
|
| 45 |
+
}
|
| 46 |
+
}
|