Upload runtime.json
Browse files- runtime.json +118 -0
runtime.json
ADDED
|
@@ -0,0 +1,118 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"format": "winstt-audio8-infinite-v1",
|
| 3 |
+
"source_model": "Edge0/Audio8-ASR-Infinite",
|
| 4 |
+
"source_revision": "7476824bc222e4ad509d286e8cae8b8d3f371129",
|
| 5 |
+
"sample_rate": 16000,
|
| 6 |
+
"samples_per_frame_len_unit": 320,
|
| 7 |
+
"look_back_samples": 840,
|
| 8 |
+
"look_ahead_samples": 40,
|
| 9 |
+
"right_pad_text_tokens": 10,
|
| 10 |
+
"left_pad_tokens_by_frame_len": {
|
| 11 |
+
"4": 18,
|
| 12 |
+
"6": 12,
|
| 13 |
+
"8": 9
|
| 14 |
+
},
|
| 15 |
+
"default_frame_len": 4,
|
| 16 |
+
"default_delay_ms": 480,
|
| 17 |
+
"encoder": {
|
| 18 |
+
"layers": 32,
|
| 19 |
+
"heads": 32,
|
| 20 |
+
"head_dim": 64,
|
| 21 |
+
"hidden": 1280,
|
| 22 |
+
"sliding_window": 750,
|
| 23 |
+
"rope_theta": 1000000.0
|
| 24 |
+
},
|
| 25 |
+
"decoder": {
|
| 26 |
+
"layers": 36,
|
| 27 |
+
"heads": 16,
|
| 28 |
+
"kv_heads": 2,
|
| 29 |
+
"head_dim": 128,
|
| 30 |
+
"hidden": 2048,
|
| 31 |
+
"rope_theta": 1000000.0,
|
| 32 |
+
"vocab": 151936
|
| 33 |
+
},
|
| 34 |
+
"rolling": {
|
| 35 |
+
"context_tokens": 375,
|
| 36 |
+
"trim_tokens": 38,
|
| 37 |
+
"stable_prefix_tokens": 16
|
| 38 |
+
},
|
| 39 |
+
"tokens": {
|
| 40 |
+
"bos": 151644,
|
| 41 |
+
"eos": 151645,
|
| 42 |
+
"pad": 151643,
|
| 43 |
+
"streaming_pad": 151665,
|
| 44 |
+
"streaming_word": 151666,
|
| 45 |
+
"language_zh": 151667,
|
| 46 |
+
"language_en": 151668
|
| 47 |
+
},
|
| 48 |
+
"semantic_vad": {
|
| 49 |
+
"horizons_seconds": [
|
| 50 |
+
0.5,
|
| 51 |
+
1.0,
|
| 52 |
+
2.0,
|
| 53 |
+
3.0
|
| 54 |
+
],
|
| 55 |
+
"num_classes": 8,
|
| 56 |
+
"end_of_turn_class": 0,
|
| 57 |
+
"default_eot_horizon_seconds": 2.0
|
| 58 |
+
},
|
| 59 |
+
"embed_tokens": {
|
| 60 |
+
"file": "embed_tokens.bf16",
|
| 61 |
+
"dtype": "bf16",
|
| 62 |
+
"shape": [
|
| 63 |
+
151936,
|
| 64 |
+
2048
|
| 65 |
+
]
|
| 66 |
+
},
|
| 67 |
+
"ada_scale": {
|
| 68 |
+
"file": "ada_scale.f32",
|
| 69 |
+
"dtype": "f32",
|
| 70 |
+
"shape": [
|
| 71 |
+
8,
|
| 72 |
+
36,
|
| 73 |
+
2048
|
| 74 |
+
],
|
| 75 |
+
"combos": [
|
| 76 |
+
{
|
| 77 |
+
"frame_len": 4,
|
| 78 |
+
"delay_ms": 240,
|
| 79 |
+
"delay_tokens": 3
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"frame_len": 4,
|
| 83 |
+
"delay_ms": 320,
|
| 84 |
+
"delay_tokens": 4
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"frame_len": 4,
|
| 88 |
+
"delay_ms": 480,
|
| 89 |
+
"delay_tokens": 6
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"frame_len": 4,
|
| 93 |
+
"delay_ms": 560,
|
| 94 |
+
"delay_tokens": 7
|
| 95 |
+
},
|
| 96 |
+
{
|
| 97 |
+
"frame_len": 6,
|
| 98 |
+
"delay_ms": 240,
|
| 99 |
+
"delay_tokens": 2
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"frame_len": 6,
|
| 103 |
+
"delay_ms": 480,
|
| 104 |
+
"delay_tokens": 4
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"frame_len": 8,
|
| 108 |
+
"delay_ms": 320,
|
| 109 |
+
"delay_tokens": 2
|
| 110 |
+
},
|
| 111 |
+
{
|
| 112 |
+
"frame_len": 8,
|
| 113 |
+
"delay_ms": 480,
|
| 114 |
+
"delay_tokens": 3
|
| 115 |
+
}
|
| 116 |
+
]
|
| 117 |
+
}
|
| 118 |
+
}
|