v3: config.json (lsd_decode_steps=16)
Browse files- onnx/config.json +3 -111
onnx/config.json
CHANGED
|
@@ -1,124 +1,16 @@
|
|
| 1 |
{
|
| 2 |
"sample_rate": 24000,
|
| 3 |
"frame_rate": 12.5,
|
| 4 |
-
"samples_per_frame": 1920,
|
| 5 |
"latent_dim": 32,
|
| 6 |
"d_model": 1024,
|
|
|
|
| 7 |
"num_heads": 16,
|
| 8 |
"head_dim": 64,
|
| 9 |
-
"
|
| 10 |
-
"
|
| 11 |
-
"flow_net_depth": 6,
|
| 12 |
-
"lsd_decode_steps": 1,
|
| 13 |
"temp": 0.4,
|
| 14 |
"eos_threshold": 0.0,
|
| 15 |
-
"vocab_size": 4001,
|
| 16 |
"rope_max_period": 100000,
|
| 17 |
-
"mimi_encoder_frame_rate": 200.0,
|
| 18 |
-
"mimi_frame_rate": 12.5,
|
| 19 |
-
"mimi_upsample_ratio": 16,
|
| 20 |
-
"bos_emb": [
|
| 21 |
-
-0.015625707805156708,
|
| 22 |
-
0.006913867779076099,
|
| 23 |
-
-0.016358541324734688,
|
| 24 |
-
0.0608685202896595,
|
| 25 |
-
0.026089003309607506,
|
| 26 |
-
0.08838313072919846,
|
| 27 |
-
-0.020075498148798943,
|
| 28 |
-
-0.0017771937418729067,
|
| 29 |
-
0.01161174476146698,
|
| 30 |
-
0.016930600628256798,
|
| 31 |
-
-0.010531092993915081,
|
| 32 |
-
-0.010811682790517807,
|
| 33 |
-
-0.05234996974468231,
|
| 34 |
-
-0.02038363367319107,
|
| 35 |
-
0.006625130772590637,
|
| 36 |
-
-0.007576636970043182,
|
| 37 |
-
0.014761571772396564,
|
| 38 |
-
-0.011585243046283722,
|
| 39 |
-
0.011751208454370499,
|
| 40 |
-
0.010635904036462307,
|
| 41 |
-
-0.008225452154874802,
|
| 42 |
-
0.05659205839037895,
|
| 43 |
-
-0.0029927159193903208,
|
| 44 |
-
-0.010195673443377018,
|
| 45 |
-
0.011196905747056007,
|
| 46 |
-
0.002476317109540105,
|
| 47 |
-
0.013769727200269699,
|
| 48 |
-
-0.012196480296552181,
|
| 49 |
-
0.0029981329571455717,
|
| 50 |
-
0.03838576376438141,
|
| 51 |
-
0.008409696631133556,
|
| 52 |
-
-0.0040644994005560875
|
| 53 |
-
],
|
| 54 |
-
"emb_mean": [
|
| 55 |
-
-0.10009765625,
|
| 56 |
-
-0.0260009765625,
|
| 57 |
-
-0.036376953125,
|
| 58 |
-
-0.341796875,
|
| 59 |
-
0.06103515625,
|
| 60 |
-
-0.80078125,
|
| 61 |
-
0.047119140625,
|
| 62 |
-
-0.0186767578125,
|
| 63 |
-
-0.10009765625,
|
| 64 |
-
-0.0693359375,
|
| 65 |
-
-0.0703125,
|
| 66 |
-
-0.1650390625,
|
| 67 |
-
0.419921875,
|
| 68 |
-
-0.0029144287109375,
|
| 69 |
-
-0.031982421875,
|
| 70 |
-
0.0673828125,
|
| 71 |
-
0.021484375,
|
| 72 |
-
-0.04443359375,
|
| 73 |
-
0.0238037109375,
|
| 74 |
-
0.0625,
|
| 75 |
-
-0.13671875,
|
| 76 |
-
-0.1787109375,
|
| 77 |
-
0.294921875,
|
| 78 |
-
0.000644683837890625,
|
| 79 |
-
-0.033203125,
|
| 80 |
-
0.11083984375,
|
| 81 |
-
0.023681640625,
|
| 82 |
-
-0.232421875,
|
| 83 |
-
-0.2060546875,
|
| 84 |
-
-0.6953125,
|
| 85 |
-
-0.05712890625,
|
| 86 |
-
0.17578125
|
| 87 |
-
],
|
| 88 |
-
"emb_std": [
|
| 89 |
-
0.96484375,
|
| 90 |
-
0.9921875,
|
| 91 |
-
1.0390625,
|
| 92 |
-
0.92578125,
|
| 93 |
-
1.015625,
|
| 94 |
-
0.59375,
|
| 95 |
-
0.98046875,
|
| 96 |
-
0.9765625,
|
| 97 |
-
0.9609375,
|
| 98 |
-
0.984375,
|
| 99 |
-
0.90625,
|
| 100 |
-
0.91015625,
|
| 101 |
-
0.9765625,
|
| 102 |
-
1.0,
|
| 103 |
-
1.0,
|
| 104 |
-
1.0,
|
| 105 |
-
1.015625,
|
| 106 |
-
0.984375,
|
| 107 |
-
1.015625,
|
| 108 |
-
1.0,
|
| 109 |
-
1.0078125,
|
| 110 |
-
1.34375,
|
| 111 |
-
0.84375,
|
| 112 |
-
1.0390625,
|
| 113 |
-
0.9453125,
|
| 114 |
-
0.96484375,
|
| 115 |
-
0.953125,
|
| 116 |
-
0.82421875,
|
| 117 |
-
0.9140625,
|
| 118 |
-
0.9453125,
|
| 119 |
-
0.9921875,
|
| 120 |
-
0.97265625
|
| 121 |
-
],
|
| 122 |
"emotions": [
|
| 123 |
"neutral",
|
| 124 |
"happy",
|
|
|
|
| 1 |
{
|
| 2 |
"sample_rate": 24000,
|
| 3 |
"frame_rate": 12.5,
|
|
|
|
| 4 |
"latent_dim": 32,
|
| 5 |
"d_model": 1024,
|
| 6 |
+
"num_layers": 6,
|
| 7 |
"num_heads": 16,
|
| 8 |
"head_dim": 64,
|
| 9 |
+
"max_cache_len": 1024,
|
| 10 |
+
"lsd_decode_steps": 16,
|
|
|
|
|
|
|
| 11 |
"temp": 0.4,
|
| 12 |
"eos_threshold": 0.0,
|
|
|
|
| 13 |
"rope_max_period": 100000,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
"emotions": [
|
| 15 |
"neutral",
|
| 16 |
"happy",
|