Upload model
Browse files- config.json +54 -9
- model.safetensors +2 -2
config.json
CHANGED
|
@@ -1,20 +1,65 @@
|
|
| 1 |
{
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
-
"
|
| 4 |
],
|
| 5 |
-
"
|
| 6 |
-
|
| 7 |
-
|
| 8 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
"kernel_size": 7,
|
| 10 |
"layer_norm_eps": 1e-06,
|
| 11 |
"layer_scale_init_value": 0.125,
|
| 12 |
-
"model_type": "
|
| 13 |
-
"n_fft":
|
| 14 |
"num_layers": 8,
|
| 15 |
"padding": 3,
|
| 16 |
-
"spec_padding": "
|
| 17 |
"torch_dtype": "float32",
|
|
|
|
| 18 |
"transformers_version": "4.54.0.dev0",
|
| 19 |
-
"use_adaptive_norm":
|
| 20 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"adanorm_num_embeddings": 4,
|
| 3 |
"architectures": [
|
| 4 |
+
"VocosWithEncodecModel"
|
| 5 |
],
|
| 6 |
+
"bandwidths": [
|
| 7 |
+
1.5,
|
| 8 |
+
3.0,
|
| 9 |
+
6.0,
|
| 10 |
+
12.0
|
| 11 |
+
],
|
| 12 |
+
"encodec_config": {
|
| 13 |
+
"audio_channels": 1,
|
| 14 |
+
"chunk_length_s": null,
|
| 15 |
+
"codebook_dim": 128,
|
| 16 |
+
"codebook_size": 1024,
|
| 17 |
+
"compress": 2,
|
| 18 |
+
"dilation_growth_rate": 2,
|
| 19 |
+
"hidden_size": 128,
|
| 20 |
+
"kernel_size": 7,
|
| 21 |
+
"last_kernel_size": 7,
|
| 22 |
+
"model_type": "encodec",
|
| 23 |
+
"norm_type": "weight_norm",
|
| 24 |
+
"normalize": false,
|
| 25 |
+
"num_filters": 32,
|
| 26 |
+
"num_lstm_layers": 2,
|
| 27 |
+
"num_residual_layers": 1,
|
| 28 |
+
"overlap": null,
|
| 29 |
+
"pad_mode": "reflect",
|
| 30 |
+
"residual_kernel_size": 3,
|
| 31 |
+
"sampling_rate": 24000,
|
| 32 |
+
"target_bandwidths": [
|
| 33 |
+
1.5,
|
| 34 |
+
3.0,
|
| 35 |
+
6.0,
|
| 36 |
+
12.0,
|
| 37 |
+
24.0
|
| 38 |
+
],
|
| 39 |
+
"trim_right_ratio": 1.0,
|
| 40 |
+
"upsampling_ratios": [
|
| 41 |
+
8,
|
| 42 |
+
5,
|
| 43 |
+
4,
|
| 44 |
+
2
|
| 45 |
+
],
|
| 46 |
+
"use_causal_conv": true,
|
| 47 |
+
"use_conv_shortcut": true
|
| 48 |
+
},
|
| 49 |
+
"hidden_dim": 384,
|
| 50 |
+
"hop_length": 320,
|
| 51 |
+
"input_channels": 128,
|
| 52 |
+
"intermediate_dim": 1152,
|
| 53 |
"kernel_size": 7,
|
| 54 |
"layer_norm_eps": 1e-06,
|
| 55 |
"layer_scale_init_value": 0.125,
|
| 56 |
+
"model_type": "vocos_with_encodec",
|
| 57 |
+
"n_fft": 1280,
|
| 58 |
"num_layers": 8,
|
| 59 |
"padding": 3,
|
| 60 |
+
"spec_padding": "same",
|
| 61 |
"torch_dtype": "float32",
|
| 62 |
+
"train_codebooks": false,
|
| 63 |
"transformers_version": "4.54.0.dev0",
|
| 64 |
+
"use_adaptive_norm": true
|
| 65 |
}
|
model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2fb9c3eba9959afd9bd012c49ecaf69bb62a2c18a250baae73236010c441fc87
|
| 3 |
+
size 133464088
|