Upload 4 files
Browse files- decoder/config.json +17 -0
- decoder/decoder_model.pt +3 -0
- encoder/config.json +22 -0
- encoder/encoder_model.pt +3 -0
decoder/config.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backbone_class": "decoder.models.VocosBackbone",
|
| 3 |
+
"head_class": "decoder.heads.ISTFTHead",
|
| 4 |
+
"backbone_config": {
|
| 5 |
+
"input_channels": 512,
|
| 6 |
+
"dim": 768,
|
| 7 |
+
"intermediate_dim": 2304,
|
| 8 |
+
"num_layers": 12,
|
| 9 |
+
"adanorm_num_embeddings": 4
|
| 10 |
+
},
|
| 11 |
+
"head_config": {
|
| 12 |
+
"dim": 768,
|
| 13 |
+
"n_fft": 1280,
|
| 14 |
+
"hop_length": 320,
|
| 15 |
+
"padding": "same"
|
| 16 |
+
}
|
| 17 |
+
}
|
decoder/decoder_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:247ad2c44c4651f2683dee953e7e6eeed672d1b96bc94e5b886e2606ff97aad1
|
| 3 |
+
size 260243154
|
encoder/config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"feature_extractor_class": "decoder.feature_extractors.EncodecFeatures",
|
| 3 |
+
"feature_extractor_config": {
|
| 4 |
+
"encodec_model": "encodec_24khz",
|
| 5 |
+
"bandwidths": [
|
| 6 |
+
6.6,
|
| 7 |
+
6.6,
|
| 8 |
+
6.6,
|
| 9 |
+
6.6
|
| 10 |
+
],
|
| 11 |
+
"train_codebooks": true,
|
| 12 |
+
"num_quantizers": 1,
|
| 13 |
+
"dowmsamples": [
|
| 14 |
+
8,
|
| 15 |
+
5,
|
| 16 |
+
4,
|
| 17 |
+
2
|
| 18 |
+
],
|
| 19 |
+
"vq_bins": 4096,
|
| 20 |
+
"vq_kmeans": 200
|
| 21 |
+
}
|
| 22 |
+
}
|
encoder/encoder_model.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:07adbb52cfc8f924ec6cc97145c8bf73da4124f8f365663d56c804fc087d3cad
|
| 3 |
+
size 87272786
|