Upload 2 files
Browse files- config.json +115 -0
- config.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"encoder_name": "WavLM",
|
| 3 |
+
"encoder_config": {
|
| 4 |
+
"hidden_dims": [
|
| 5 |
+
512,
|
| 6 |
+
512,
|
| 7 |
+
512,
|
| 8 |
+
512,
|
| 9 |
+
512,
|
| 10 |
+
512,
|
| 11 |
+
512
|
| 12 |
+
],
|
| 13 |
+
"kernel_sizes": [
|
| 14 |
+
10,
|
| 15 |
+
3,
|
| 16 |
+
3,
|
| 17 |
+
3,
|
| 18 |
+
3,
|
| 19 |
+
2,
|
| 20 |
+
2
|
| 21 |
+
],
|
| 22 |
+
"strides": [
|
| 23 |
+
5,
|
| 24 |
+
2,
|
| 25 |
+
2,
|
| 26 |
+
2,
|
| 27 |
+
2,
|
| 28 |
+
2,
|
| 29 |
+
2
|
| 30 |
+
],
|
| 31 |
+
"num_layers": 6,
|
| 32 |
+
"dim": 1024,
|
| 33 |
+
"ffn_dim": 4096,
|
| 34 |
+
"num_heads": 16,
|
| 35 |
+
"num_buckets": 320,
|
| 36 |
+
"max_distance": 800,
|
| 37 |
+
"max_cached_steps": 2048,
|
| 38 |
+
"dropout": 0.0,
|
| 39 |
+
"conv_pos": 128,
|
| 40 |
+
"conv_pos_groups": 16,
|
| 41 |
+
"causal": true,
|
| 42 |
+
"window_size": 512,
|
| 43 |
+
"lookahead_size": 3,
|
| 44 |
+
"use_flex_attention": false
|
| 45 |
+
},
|
| 46 |
+
"compressor_name": "FocalEncoder",
|
| 47 |
+
"compressor_config": {
|
| 48 |
+
"input_dim": 1024,
|
| 49 |
+
"output_dim": 16,
|
| 50 |
+
"hidden_dims": [
|
| 51 |
+
1024,
|
| 52 |
+
1024,
|
| 53 |
+
1024
|
| 54 |
+
],
|
| 55 |
+
"downscale_factors": [
|
| 56 |
+
1,
|
| 57 |
+
1,
|
| 58 |
+
1
|
| 59 |
+
],
|
| 60 |
+
"focal_window": 14,
|
| 61 |
+
"focal_level": 2,
|
| 62 |
+
"focal_factor": 4,
|
| 63 |
+
"dropout": 0.0,
|
| 64 |
+
"use_post_norm": false,
|
| 65 |
+
"use_layerscale": false,
|
| 66 |
+
"layerscale_init": 0.0001,
|
| 67 |
+
"tanhscale_init": 0.5,
|
| 68 |
+
"normalize_modulator": false,
|
| 69 |
+
"causal": true,
|
| 70 |
+
"window_size": 512
|
| 71 |
+
},
|
| 72 |
+
"quantizer_name": "BinarySphericalQuantizer",
|
| 73 |
+
"quantizer_config": {
|
| 74 |
+
"codebook_size": 65536
|
| 75 |
+
},
|
| 76 |
+
"decompressor_name": "FocalDecoder",
|
| 77 |
+
"decompressor_config": {
|
| 78 |
+
"input_dim": 16,
|
| 79 |
+
"output_dim": 1024,
|
| 80 |
+
"hidden_dims": [
|
| 81 |
+
1024,
|
| 82 |
+
1024,
|
| 83 |
+
1024
|
| 84 |
+
],
|
| 85 |
+
"upscale_factors": [
|
| 86 |
+
1,
|
| 87 |
+
1,
|
| 88 |
+
1
|
| 89 |
+
],
|
| 90 |
+
"focal_window": 14,
|
| 91 |
+
"focal_level": 2,
|
| 92 |
+
"focal_factor": 4,
|
| 93 |
+
"dropout": 0.0,
|
| 94 |
+
"use_post_norm": false,
|
| 95 |
+
"use_layerscale": false,
|
| 96 |
+
"layerscale_init": 0.0001,
|
| 97 |
+
"tanhscale_init": 0.5,
|
| 98 |
+
"normalize_modulator": false,
|
| 99 |
+
"causal": true,
|
| 100 |
+
"window_size": 512,
|
| 101 |
+
"last_window_size": 512,
|
| 102 |
+
"lookahead_size": 3
|
| 103 |
+
},
|
| 104 |
+
"decoder_name": "WaveNeXt",
|
| 105 |
+
"decoder_config": {
|
| 106 |
+
"input_dim": 1024,
|
| 107 |
+
"num_layers": 8,
|
| 108 |
+
"dim": 1024,
|
| 109 |
+
"ffn_dim": 2048,
|
| 110 |
+
"kernel_size": 7,
|
| 111 |
+
"layerscale_init": null,
|
| 112 |
+
"hop_length": 480,
|
| 113 |
+
"causal": true
|
| 114 |
+
}
|
| 115 |
+
}
|
config.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:81d56de414207a59555b69536eaf39f7d70fce3193439e1617f4bfad18d2787b
|
| 3 |
+
size 996226872
|