Upload folder using huggingface_hub
Browse files- FLowHigh_basic_400k.json +45 -0
- FLowHigh_basic_400k.pt +3 -0
- bigvgan_48khz_256band.json +45 -0
- bigvgan_48khz_256band.pt +3 -0
FLowHigh_basic_400k.json
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"random_seed": 104,
|
3 |
+
"data": {
|
4 |
+
"data_path": "The/path/to/train/directory/path",
|
5 |
+
"valid_path": "The/path/to/test/directory/path",
|
6 |
+
"valid_prepare": true,
|
7 |
+
"samplingrate": 48000,
|
8 |
+
"max_wav_value": 32767.0,
|
9 |
+
"n_fft": 2048,
|
10 |
+
"hop_length": 480,
|
11 |
+
"win_length": 2048,
|
12 |
+
"n_mel_channels": 256,
|
13 |
+
"mel_fmin": 20,
|
14 |
+
"mel_fmax": 24000,
|
15 |
+
"downsample_min": 4000,
|
16 |
+
"downsample_max": 32000,
|
17 |
+
"downsampling_method": "scipy"
|
18 |
+
},
|
19 |
+
"model": {
|
20 |
+
"modelname": "FLowHigh",
|
21 |
+
"architecture": "transformer",
|
22 |
+
"dim": 1024,
|
23 |
+
"n_layers": 2,
|
24 |
+
"n_heads": 16,
|
25 |
+
"dim_head":64,
|
26 |
+
"cfm_path": "basic_cfm",
|
27 |
+
"sigma":1e-4,
|
28 |
+
"vocoder": "bigvgan",
|
29 |
+
"vocoderpath": "/<Path>/vocoder/BIGVGAN/checkpoint/g_48_00850000",
|
30 |
+
"vocoderconfigpath": "/<Path>/vocoder/BIGVGAN/config/bigvgan_48khz_256band_config.json"
|
31 |
+
},
|
32 |
+
"train": {
|
33 |
+
"random_split_seed": 53,
|
34 |
+
"batchsize": 128,
|
35 |
+
"lr": 3e-4,
|
36 |
+
"initial_lr": 1e-5,
|
37 |
+
"n_train_steps": 400001,
|
38 |
+
"n_warmup_steps": 0,
|
39 |
+
"log_every": 10000,
|
40 |
+
"save_results_every": 10000,
|
41 |
+
"save_model_every": 100000,
|
42 |
+
"save_dir": "The/path/to/save/FLowHigh/",
|
43 |
+
"weighted_loss": false
|
44 |
+
}
|
45 |
+
}
|
FLowHigh_basic_400k.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:84688e90d09b6f0788aeabc351ebc5f8d86463adb7ecf7ca3ef9c548e3825b8f
|
3 |
+
size 481490826
|
bigvgan_48khz_256band.json
ADDED
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"resblock": "1",
|
3 |
+
"num_gpus": 0,
|
4 |
+
"batch_size": 64,
|
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": [6,5,4,2,2],
|
12 |
+
"upsample_kernel_sizes": [12,11,8,4,4],
|
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 |
+
"resolutions": [[1024, 120, 600], [2048, 240, 1200], [512, 50, 240]],
|
21 |
+
"mpd_reshapes": [2, 3, 5, 7, 11],
|
22 |
+
"use_spectral_norm": false,
|
23 |
+
"discriminator_channel_mult": 1,
|
24 |
+
|
25 |
+
"segment_size": 15360,
|
26 |
+
"num_mels": 256,
|
27 |
+
"num_freq": 2049,
|
28 |
+
"n_fft": 2048,
|
29 |
+
"hop_size": 480,
|
30 |
+
"win_size": 2048,
|
31 |
+
|
32 |
+
"sampling_rate": 48000,
|
33 |
+
|
34 |
+
"fmin": 20,
|
35 |
+
"fmax": 24000,
|
36 |
+
"fmax_for_loss": null,
|
37 |
+
|
38 |
+
"num_workers": 4,
|
39 |
+
|
40 |
+
"dist_config": {
|
41 |
+
"dist_backend": "nccl",
|
42 |
+
"dist_url": "tcp://localhost:54321",
|
43 |
+
"world_size": 1
|
44 |
+
}
|
45 |
+
}
|
bigvgan_48khz_256band.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:40c9fbe33e8d9f4090b988733996984e899c2ba69e475435e49704c8378c14bb
|
3 |
+
size 56105238
|