Upload folder using huggingface_hub
Browse files- pocket_ko/bos_before_voice.npy +3 -0
- pocket_ko/bundle.json +879 -0
- pocket_ko/default_voice_p5.safetensors +3 -0
- pocket_ko/flow_lm_flow_int8.onnx +3 -0
- pocket_ko/flow_lm_main_int8.onnx +3 -0
- pocket_ko/mimi_decoder.onnx +3 -0
- pocket_ko/mimi_encoder.onnx +3 -0
- pocket_ko/model_manifest.json +38 -0
- pocket_ko/text_conditioner.onnx +3 -0
- pocket_ko/tokenizer.model +3 -0
pocket_ko/bos_before_voice.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e2400fb47b3f5d45ece082e5d563b4d8e87ee6d4b72782b8fbb432b2e5a6ae4d
|
| 3 |
+
size 4224
|
pocket_ko/bundle.json
ADDED
|
@@ -0,0 +1,879 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"schema_version": 2,
|
| 3 |
+
"language": "korean_stock100m",
|
| 4 |
+
"bundle_name": "stock100m",
|
| 5 |
+
"sample_rate": 24000,
|
| 6 |
+
"frame_rate": 12.5,
|
| 7 |
+
"samples_per_frame": 1920,
|
| 8 |
+
"latent_dim": 32,
|
| 9 |
+
"conditioning_dim": 1024,
|
| 10 |
+
"insert_bos_before_voice": true,
|
| 11 |
+
"bos_before_voice_file": "bos_before_voice.npy",
|
| 12 |
+
"max_token_per_chunk": 50,
|
| 13 |
+
"remove_semicolons": false,
|
| 14 |
+
"pad_with_spaces_for_short_inputs": false,
|
| 15 |
+
"model_recommended_frames_after_eos": null,
|
| 16 |
+
"predefined_voices": [],
|
| 17 |
+
"tokenizer_file": "tokenizer.model",
|
| 18 |
+
"flow_lm_state_manifest": [
|
| 19 |
+
{
|
| 20 |
+
"module": "transformer.layers.0.self_attn",
|
| 21 |
+
"key": "cache",
|
| 22 |
+
"dtype": "float32",
|
| 23 |
+
"fill": "nan",
|
| 24 |
+
"shape": [
|
| 25 |
+
2,
|
| 26 |
+
1,
|
| 27 |
+
1000,
|
| 28 |
+
16,
|
| 29 |
+
64
|
| 30 |
+
],
|
| 31 |
+
"index": 0,
|
| 32 |
+
"input_name": "state_0",
|
| 33 |
+
"output_name": "out_state_0"
|
| 34 |
+
},
|
| 35 |
+
{
|
| 36 |
+
"module": "transformer.layers.0.self_attn",
|
| 37 |
+
"key": "offset",
|
| 38 |
+
"dtype": "int64",
|
| 39 |
+
"fill": "zeros",
|
| 40 |
+
"shape": [
|
| 41 |
+
1
|
| 42 |
+
],
|
| 43 |
+
"index": 1,
|
| 44 |
+
"input_name": "state_1",
|
| 45 |
+
"output_name": "out_state_1"
|
| 46 |
+
},
|
| 47 |
+
{
|
| 48 |
+
"module": "transformer.layers.1.self_attn",
|
| 49 |
+
"key": "cache",
|
| 50 |
+
"dtype": "float32",
|
| 51 |
+
"fill": "nan",
|
| 52 |
+
"shape": [
|
| 53 |
+
2,
|
| 54 |
+
1,
|
| 55 |
+
1000,
|
| 56 |
+
16,
|
| 57 |
+
64
|
| 58 |
+
],
|
| 59 |
+
"index": 2,
|
| 60 |
+
"input_name": "state_2",
|
| 61 |
+
"output_name": "out_state_2"
|
| 62 |
+
},
|
| 63 |
+
{
|
| 64 |
+
"module": "transformer.layers.1.self_attn",
|
| 65 |
+
"key": "offset",
|
| 66 |
+
"dtype": "int64",
|
| 67 |
+
"fill": "zeros",
|
| 68 |
+
"shape": [
|
| 69 |
+
1
|
| 70 |
+
],
|
| 71 |
+
"index": 3,
|
| 72 |
+
"input_name": "state_3",
|
| 73 |
+
"output_name": "out_state_3"
|
| 74 |
+
},
|
| 75 |
+
{
|
| 76 |
+
"module": "transformer.layers.2.self_attn",
|
| 77 |
+
"key": "cache",
|
| 78 |
+
"dtype": "float32",
|
| 79 |
+
"fill": "nan",
|
| 80 |
+
"shape": [
|
| 81 |
+
2,
|
| 82 |
+
1,
|
| 83 |
+
1000,
|
| 84 |
+
16,
|
| 85 |
+
64
|
| 86 |
+
],
|
| 87 |
+
"index": 4,
|
| 88 |
+
"input_name": "state_4",
|
| 89 |
+
"output_name": "out_state_4"
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"module": "transformer.layers.2.self_attn",
|
| 93 |
+
"key": "offset",
|
| 94 |
+
"dtype": "int64",
|
| 95 |
+
"fill": "zeros",
|
| 96 |
+
"shape": [
|
| 97 |
+
1
|
| 98 |
+
],
|
| 99 |
+
"index": 5,
|
| 100 |
+
"input_name": "state_5",
|
| 101 |
+
"output_name": "out_state_5"
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"module": "transformer.layers.3.self_attn",
|
| 105 |
+
"key": "cache",
|
| 106 |
+
"dtype": "float32",
|
| 107 |
+
"fill": "nan",
|
| 108 |
+
"shape": [
|
| 109 |
+
2,
|
| 110 |
+
1,
|
| 111 |
+
1000,
|
| 112 |
+
16,
|
| 113 |
+
64
|
| 114 |
+
],
|
| 115 |
+
"index": 6,
|
| 116 |
+
"input_name": "state_6",
|
| 117 |
+
"output_name": "out_state_6"
|
| 118 |
+
},
|
| 119 |
+
{
|
| 120 |
+
"module": "transformer.layers.3.self_attn",
|
| 121 |
+
"key": "offset",
|
| 122 |
+
"dtype": "int64",
|
| 123 |
+
"fill": "zeros",
|
| 124 |
+
"shape": [
|
| 125 |
+
1
|
| 126 |
+
],
|
| 127 |
+
"index": 7,
|
| 128 |
+
"input_name": "state_7",
|
| 129 |
+
"output_name": "out_state_7"
|
| 130 |
+
},
|
| 131 |
+
{
|
| 132 |
+
"module": "transformer.layers.4.self_attn",
|
| 133 |
+
"key": "cache",
|
| 134 |
+
"dtype": "float32",
|
| 135 |
+
"fill": "nan",
|
| 136 |
+
"shape": [
|
| 137 |
+
2,
|
| 138 |
+
1,
|
| 139 |
+
1000,
|
| 140 |
+
16,
|
| 141 |
+
64
|
| 142 |
+
],
|
| 143 |
+
"index": 8,
|
| 144 |
+
"input_name": "state_8",
|
| 145 |
+
"output_name": "out_state_8"
|
| 146 |
+
},
|
| 147 |
+
{
|
| 148 |
+
"module": "transformer.layers.4.self_attn",
|
| 149 |
+
"key": "offset",
|
| 150 |
+
"dtype": "int64",
|
| 151 |
+
"fill": "zeros",
|
| 152 |
+
"shape": [
|
| 153 |
+
1
|
| 154 |
+
],
|
| 155 |
+
"index": 9,
|
| 156 |
+
"input_name": "state_9",
|
| 157 |
+
"output_name": "out_state_9"
|
| 158 |
+
},
|
| 159 |
+
{
|
| 160 |
+
"module": "transformer.layers.5.self_attn",
|
| 161 |
+
"key": "cache",
|
| 162 |
+
"dtype": "float32",
|
| 163 |
+
"fill": "nan",
|
| 164 |
+
"shape": [
|
| 165 |
+
2,
|
| 166 |
+
1,
|
| 167 |
+
1000,
|
| 168 |
+
16,
|
| 169 |
+
64
|
| 170 |
+
],
|
| 171 |
+
"index": 10,
|
| 172 |
+
"input_name": "state_10",
|
| 173 |
+
"output_name": "out_state_10"
|
| 174 |
+
},
|
| 175 |
+
{
|
| 176 |
+
"module": "transformer.layers.5.self_attn",
|
| 177 |
+
"key": "offset",
|
| 178 |
+
"dtype": "int64",
|
| 179 |
+
"fill": "zeros",
|
| 180 |
+
"shape": [
|
| 181 |
+
1
|
| 182 |
+
],
|
| 183 |
+
"index": 11,
|
| 184 |
+
"input_name": "state_11",
|
| 185 |
+
"output_name": "out_state_11"
|
| 186 |
+
}
|
| 187 |
+
],
|
| 188 |
+
"mimi_state_manifest": [
|
| 189 |
+
{
|
| 190 |
+
"module": "decoder.model.0",
|
| 191 |
+
"key": "first",
|
| 192 |
+
"dtype": "bool",
|
| 193 |
+
"fill": "ones",
|
| 194 |
+
"shape": [
|
| 195 |
+
1
|
| 196 |
+
],
|
| 197 |
+
"index": 0,
|
| 198 |
+
"input_name": "state_0",
|
| 199 |
+
"output_name": "out_state_0"
|
| 200 |
+
},
|
| 201 |
+
{
|
| 202 |
+
"module": "decoder.model.0",
|
| 203 |
+
"key": "previous",
|
| 204 |
+
"dtype": "float32",
|
| 205 |
+
"fill": "zeros",
|
| 206 |
+
"shape": [
|
| 207 |
+
1,
|
| 208 |
+
512,
|
| 209 |
+
6
|
| 210 |
+
],
|
| 211 |
+
"index": 1,
|
| 212 |
+
"input_name": "state_1",
|
| 213 |
+
"output_name": "out_state_1"
|
| 214 |
+
},
|
| 215 |
+
{
|
| 216 |
+
"module": "decoder.model.11",
|
| 217 |
+
"key": "first",
|
| 218 |
+
"dtype": "bool",
|
| 219 |
+
"fill": "ones",
|
| 220 |
+
"shape": [
|
| 221 |
+
1
|
| 222 |
+
],
|
| 223 |
+
"index": 2,
|
| 224 |
+
"input_name": "state_2",
|
| 225 |
+
"output_name": "out_state_2"
|
| 226 |
+
},
|
| 227 |
+
{
|
| 228 |
+
"module": "decoder.model.11",
|
| 229 |
+
"key": "previous",
|
| 230 |
+
"dtype": "float32",
|
| 231 |
+
"fill": "zeros",
|
| 232 |
+
"shape": [
|
| 233 |
+
1,
|
| 234 |
+
64,
|
| 235 |
+
2
|
| 236 |
+
],
|
| 237 |
+
"index": 3,
|
| 238 |
+
"input_name": "state_3",
|
| 239 |
+
"output_name": "out_state_3"
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"module": "decoder.model.2",
|
| 243 |
+
"key": "partial",
|
| 244 |
+
"dtype": "float32",
|
| 245 |
+
"fill": "zeros",
|
| 246 |
+
"shape": [
|
| 247 |
+
1,
|
| 248 |
+
256,
|
| 249 |
+
6
|
| 250 |
+
],
|
| 251 |
+
"index": 4,
|
| 252 |
+
"input_name": "state_4",
|
| 253 |
+
"output_name": "out_state_4"
|
| 254 |
+
},
|
| 255 |
+
{
|
| 256 |
+
"module": "decoder.model.3.block.1",
|
| 257 |
+
"key": "first",
|
| 258 |
+
"dtype": "bool",
|
| 259 |
+
"fill": "ones",
|
| 260 |
+
"shape": [
|
| 261 |
+
1
|
| 262 |
+
],
|
| 263 |
+
"index": 5,
|
| 264 |
+
"input_name": "state_5",
|
| 265 |
+
"output_name": "out_state_5"
|
| 266 |
+
},
|
| 267 |
+
{
|
| 268 |
+
"module": "decoder.model.3.block.1",
|
| 269 |
+
"key": "previous",
|
| 270 |
+
"dtype": "float32",
|
| 271 |
+
"fill": "zeros",
|
| 272 |
+
"shape": [
|
| 273 |
+
1,
|
| 274 |
+
256,
|
| 275 |
+
2
|
| 276 |
+
],
|
| 277 |
+
"index": 6,
|
| 278 |
+
"input_name": "state_6",
|
| 279 |
+
"output_name": "out_state_6"
|
| 280 |
+
},
|
| 281 |
+
{
|
| 282 |
+
"module": "decoder.model.3.block.3",
|
| 283 |
+
"key": "first",
|
| 284 |
+
"dtype": "bool",
|
| 285 |
+
"fill": "ones",
|
| 286 |
+
"shape": [
|
| 287 |
+
1
|
| 288 |
+
],
|
| 289 |
+
"index": 7,
|
| 290 |
+
"input_name": "state_7",
|
| 291 |
+
"output_name": "out_state_7"
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"module": "decoder.model.3.block.3",
|
| 295 |
+
"key": "previous",
|
| 296 |
+
"dtype": "float32",
|
| 297 |
+
"fill": "zeros",
|
| 298 |
+
"shape": [
|
| 299 |
+
1,
|
| 300 |
+
128,
|
| 301 |
+
0
|
| 302 |
+
],
|
| 303 |
+
"index": 8,
|
| 304 |
+
"input_name": "state_8",
|
| 305 |
+
"output_name": "out_state_8"
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"module": "decoder.model.5",
|
| 309 |
+
"key": "partial",
|
| 310 |
+
"dtype": "float32",
|
| 311 |
+
"fill": "zeros",
|
| 312 |
+
"shape": [
|
| 313 |
+
1,
|
| 314 |
+
128,
|
| 315 |
+
5
|
| 316 |
+
],
|
| 317 |
+
"index": 9,
|
| 318 |
+
"input_name": "state_9",
|
| 319 |
+
"output_name": "out_state_9"
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"module": "decoder.model.6.block.1",
|
| 323 |
+
"key": "first",
|
| 324 |
+
"dtype": "bool",
|
| 325 |
+
"fill": "ones",
|
| 326 |
+
"shape": [
|
| 327 |
+
1
|
| 328 |
+
],
|
| 329 |
+
"index": 10,
|
| 330 |
+
"input_name": "state_10",
|
| 331 |
+
"output_name": "out_state_10"
|
| 332 |
+
},
|
| 333 |
+
{
|
| 334 |
+
"module": "decoder.model.6.block.1",
|
| 335 |
+
"key": "previous",
|
| 336 |
+
"dtype": "float32",
|
| 337 |
+
"fill": "zeros",
|
| 338 |
+
"shape": [
|
| 339 |
+
1,
|
| 340 |
+
128,
|
| 341 |
+
2
|
| 342 |
+
],
|
| 343 |
+
"index": 11,
|
| 344 |
+
"input_name": "state_11",
|
| 345 |
+
"output_name": "out_state_11"
|
| 346 |
+
},
|
| 347 |
+
{
|
| 348 |
+
"module": "decoder.model.6.block.3",
|
| 349 |
+
"key": "first",
|
| 350 |
+
"dtype": "bool",
|
| 351 |
+
"fill": "ones",
|
| 352 |
+
"shape": [
|
| 353 |
+
1
|
| 354 |
+
],
|
| 355 |
+
"index": 12,
|
| 356 |
+
"input_name": "state_12",
|
| 357 |
+
"output_name": "out_state_12"
|
| 358 |
+
},
|
| 359 |
+
{
|
| 360 |
+
"module": "decoder.model.6.block.3",
|
| 361 |
+
"key": "previous",
|
| 362 |
+
"dtype": "float32",
|
| 363 |
+
"fill": "zeros",
|
| 364 |
+
"shape": [
|
| 365 |
+
1,
|
| 366 |
+
64,
|
| 367 |
+
0
|
| 368 |
+
],
|
| 369 |
+
"index": 13,
|
| 370 |
+
"input_name": "state_13",
|
| 371 |
+
"output_name": "out_state_13"
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"module": "decoder.model.8",
|
| 375 |
+
"key": "partial",
|
| 376 |
+
"dtype": "float32",
|
| 377 |
+
"fill": "zeros",
|
| 378 |
+
"shape": [
|
| 379 |
+
1,
|
| 380 |
+
64,
|
| 381 |
+
4
|
| 382 |
+
],
|
| 383 |
+
"index": 14,
|
| 384 |
+
"input_name": "state_14",
|
| 385 |
+
"output_name": "out_state_14"
|
| 386 |
+
},
|
| 387 |
+
{
|
| 388 |
+
"module": "decoder.model.9.block.1",
|
| 389 |
+
"key": "first",
|
| 390 |
+
"dtype": "bool",
|
| 391 |
+
"fill": "ones",
|
| 392 |
+
"shape": [
|
| 393 |
+
1
|
| 394 |
+
],
|
| 395 |
+
"index": 15,
|
| 396 |
+
"input_name": "state_15",
|
| 397 |
+
"output_name": "out_state_15"
|
| 398 |
+
},
|
| 399 |
+
{
|
| 400 |
+
"module": "decoder.model.9.block.1",
|
| 401 |
+
"key": "previous",
|
| 402 |
+
"dtype": "float32",
|
| 403 |
+
"fill": "zeros",
|
| 404 |
+
"shape": [
|
| 405 |
+
1,
|
| 406 |
+
64,
|
| 407 |
+
2
|
| 408 |
+
],
|
| 409 |
+
"index": 16,
|
| 410 |
+
"input_name": "state_16",
|
| 411 |
+
"output_name": "out_state_16"
|
| 412 |
+
},
|
| 413 |
+
{
|
| 414 |
+
"module": "decoder.model.9.block.3",
|
| 415 |
+
"key": "first",
|
| 416 |
+
"dtype": "bool",
|
| 417 |
+
"fill": "ones",
|
| 418 |
+
"shape": [
|
| 419 |
+
1
|
| 420 |
+
],
|
| 421 |
+
"index": 17,
|
| 422 |
+
"input_name": "state_17",
|
| 423 |
+
"output_name": "out_state_17"
|
| 424 |
+
},
|
| 425 |
+
{
|
| 426 |
+
"module": "decoder.model.9.block.3",
|
| 427 |
+
"key": "previous",
|
| 428 |
+
"dtype": "float32",
|
| 429 |
+
"fill": "zeros",
|
| 430 |
+
"shape": [
|
| 431 |
+
1,
|
| 432 |
+
32,
|
| 433 |
+
0
|
| 434 |
+
],
|
| 435 |
+
"index": 18,
|
| 436 |
+
"input_name": "state_18",
|
| 437 |
+
"output_name": "out_state_18"
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"module": "decoder_transformer.transformer.layers.0.self_attn",
|
| 441 |
+
"key": "cache",
|
| 442 |
+
"dtype": "float32",
|
| 443 |
+
"fill": "nan",
|
| 444 |
+
"shape": [
|
| 445 |
+
2,
|
| 446 |
+
1,
|
| 447 |
+
8192,
|
| 448 |
+
8,
|
| 449 |
+
64
|
| 450 |
+
],
|
| 451 |
+
"index": 19,
|
| 452 |
+
"input_name": "state_19",
|
| 453 |
+
"output_name": "out_state_19"
|
| 454 |
+
},
|
| 455 |
+
{
|
| 456 |
+
"module": "decoder_transformer.transformer.layers.0.self_attn",
|
| 457 |
+
"key": "offset",
|
| 458 |
+
"dtype": "int64",
|
| 459 |
+
"fill": "zeros",
|
| 460 |
+
"shape": [
|
| 461 |
+
1
|
| 462 |
+
],
|
| 463 |
+
"index": 20,
|
| 464 |
+
"input_name": "state_20",
|
| 465 |
+
"output_name": "out_state_20"
|
| 466 |
+
},
|
| 467 |
+
{
|
| 468 |
+
"module": "decoder_transformer.transformer.layers.1.self_attn",
|
| 469 |
+
"key": "cache",
|
| 470 |
+
"dtype": "float32",
|
| 471 |
+
"fill": "nan",
|
| 472 |
+
"shape": [
|
| 473 |
+
2,
|
| 474 |
+
1,
|
| 475 |
+
8192,
|
| 476 |
+
8,
|
| 477 |
+
64
|
| 478 |
+
],
|
| 479 |
+
"index": 21,
|
| 480 |
+
"input_name": "state_21",
|
| 481 |
+
"output_name": "out_state_21"
|
| 482 |
+
},
|
| 483 |
+
{
|
| 484 |
+
"module": "decoder_transformer.transformer.layers.1.self_attn",
|
| 485 |
+
"key": "offset",
|
| 486 |
+
"dtype": "int64",
|
| 487 |
+
"fill": "zeros",
|
| 488 |
+
"shape": [
|
| 489 |
+
1
|
| 490 |
+
],
|
| 491 |
+
"index": 22,
|
| 492 |
+
"input_name": "state_22",
|
| 493 |
+
"output_name": "out_state_22"
|
| 494 |
+
},
|
| 495 |
+
{
|
| 496 |
+
"module": "downsample.conv",
|
| 497 |
+
"key": "first",
|
| 498 |
+
"dtype": "bool",
|
| 499 |
+
"fill": "ones",
|
| 500 |
+
"shape": [
|
| 501 |
+
1
|
| 502 |
+
],
|
| 503 |
+
"index": 23,
|
| 504 |
+
"input_name": "state_23",
|
| 505 |
+
"output_name": "out_state_23"
|
| 506 |
+
},
|
| 507 |
+
{
|
| 508 |
+
"module": "downsample.conv",
|
| 509 |
+
"key": "previous",
|
| 510 |
+
"dtype": "float32",
|
| 511 |
+
"fill": "zeros",
|
| 512 |
+
"shape": [
|
| 513 |
+
1,
|
| 514 |
+
512,
|
| 515 |
+
16
|
| 516 |
+
],
|
| 517 |
+
"index": 24,
|
| 518 |
+
"input_name": "state_24",
|
| 519 |
+
"output_name": "out_state_24"
|
| 520 |
+
},
|
| 521 |
+
{
|
| 522 |
+
"module": "encoder.model.0",
|
| 523 |
+
"key": "first",
|
| 524 |
+
"dtype": "bool",
|
| 525 |
+
"fill": "ones",
|
| 526 |
+
"shape": [
|
| 527 |
+
1
|
| 528 |
+
],
|
| 529 |
+
"index": 25,
|
| 530 |
+
"input_name": "state_25",
|
| 531 |
+
"output_name": "out_state_25"
|
| 532 |
+
},
|
| 533 |
+
{
|
| 534 |
+
"module": "encoder.model.0",
|
| 535 |
+
"key": "previous",
|
| 536 |
+
"dtype": "float32",
|
| 537 |
+
"fill": "zeros",
|
| 538 |
+
"shape": [
|
| 539 |
+
1,
|
| 540 |
+
1,
|
| 541 |
+
6
|
| 542 |
+
],
|
| 543 |
+
"index": 26,
|
| 544 |
+
"input_name": "state_26",
|
| 545 |
+
"output_name": "out_state_26"
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"module": "encoder.model.1.block.1",
|
| 549 |
+
"key": "first",
|
| 550 |
+
"dtype": "bool",
|
| 551 |
+
"fill": "ones",
|
| 552 |
+
"shape": [
|
| 553 |
+
1
|
| 554 |
+
],
|
| 555 |
+
"index": 27,
|
| 556 |
+
"input_name": "state_27",
|
| 557 |
+
"output_name": "out_state_27"
|
| 558 |
+
},
|
| 559 |
+
{
|
| 560 |
+
"module": "encoder.model.1.block.1",
|
| 561 |
+
"key": "previous",
|
| 562 |
+
"dtype": "float32",
|
| 563 |
+
"fill": "zeros",
|
| 564 |
+
"shape": [
|
| 565 |
+
1,
|
| 566 |
+
64,
|
| 567 |
+
2
|
| 568 |
+
],
|
| 569 |
+
"index": 28,
|
| 570 |
+
"input_name": "state_28",
|
| 571 |
+
"output_name": "out_state_28"
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"module": "encoder.model.1.block.3",
|
| 575 |
+
"key": "first",
|
| 576 |
+
"dtype": "bool",
|
| 577 |
+
"fill": "ones",
|
| 578 |
+
"shape": [
|
| 579 |
+
1
|
| 580 |
+
],
|
| 581 |
+
"index": 29,
|
| 582 |
+
"input_name": "state_29",
|
| 583 |
+
"output_name": "out_state_29"
|
| 584 |
+
},
|
| 585 |
+
{
|
| 586 |
+
"module": "encoder.model.1.block.3",
|
| 587 |
+
"key": "previous",
|
| 588 |
+
"dtype": "float32",
|
| 589 |
+
"fill": "zeros",
|
| 590 |
+
"shape": [
|
| 591 |
+
1,
|
| 592 |
+
32,
|
| 593 |
+
0
|
| 594 |
+
],
|
| 595 |
+
"index": 30,
|
| 596 |
+
"input_name": "state_30",
|
| 597 |
+
"output_name": "out_state_30"
|
| 598 |
+
},
|
| 599 |
+
{
|
| 600 |
+
"module": "encoder.model.11",
|
| 601 |
+
"key": "first",
|
| 602 |
+
"dtype": "bool",
|
| 603 |
+
"fill": "ones",
|
| 604 |
+
"shape": [
|
| 605 |
+
1
|
| 606 |
+
],
|
| 607 |
+
"index": 31,
|
| 608 |
+
"input_name": "state_31",
|
| 609 |
+
"output_name": "out_state_31"
|
| 610 |
+
},
|
| 611 |
+
{
|
| 612 |
+
"module": "encoder.model.11",
|
| 613 |
+
"key": "previous",
|
| 614 |
+
"dtype": "float32",
|
| 615 |
+
"fill": "zeros",
|
| 616 |
+
"shape": [
|
| 617 |
+
1,
|
| 618 |
+
512,
|
| 619 |
+
2
|
| 620 |
+
],
|
| 621 |
+
"index": 32,
|
| 622 |
+
"input_name": "state_32",
|
| 623 |
+
"output_name": "out_state_32"
|
| 624 |
+
},
|
| 625 |
+
{
|
| 626 |
+
"module": "encoder.model.3",
|
| 627 |
+
"key": "first",
|
| 628 |
+
"dtype": "bool",
|
| 629 |
+
"fill": "ones",
|
| 630 |
+
"shape": [
|
| 631 |
+
1
|
| 632 |
+
],
|
| 633 |
+
"index": 33,
|
| 634 |
+
"input_name": "state_33",
|
| 635 |
+
"output_name": "out_state_33"
|
| 636 |
+
},
|
| 637 |
+
{
|
| 638 |
+
"module": "encoder.model.3",
|
| 639 |
+
"key": "previous",
|
| 640 |
+
"dtype": "float32",
|
| 641 |
+
"fill": "zeros",
|
| 642 |
+
"shape": [
|
| 643 |
+
1,
|
| 644 |
+
64,
|
| 645 |
+
4
|
| 646 |
+
],
|
| 647 |
+
"index": 34,
|
| 648 |
+
"input_name": "state_34",
|
| 649 |
+
"output_name": "out_state_34"
|
| 650 |
+
},
|
| 651 |
+
{
|
| 652 |
+
"module": "encoder.model.4.block.1",
|
| 653 |
+
"key": "first",
|
| 654 |
+
"dtype": "bool",
|
| 655 |
+
"fill": "ones",
|
| 656 |
+
"shape": [
|
| 657 |
+
1
|
| 658 |
+
],
|
| 659 |
+
"index": 35,
|
| 660 |
+
"input_name": "state_35",
|
| 661 |
+
"output_name": "out_state_35"
|
| 662 |
+
},
|
| 663 |
+
{
|
| 664 |
+
"module": "encoder.model.4.block.1",
|
| 665 |
+
"key": "previous",
|
| 666 |
+
"dtype": "float32",
|
| 667 |
+
"fill": "zeros",
|
| 668 |
+
"shape": [
|
| 669 |
+
1,
|
| 670 |
+
128,
|
| 671 |
+
2
|
| 672 |
+
],
|
| 673 |
+
"index": 36,
|
| 674 |
+
"input_name": "state_36",
|
| 675 |
+
"output_name": "out_state_36"
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"module": "encoder.model.4.block.3",
|
| 679 |
+
"key": "first",
|
| 680 |
+
"dtype": "bool",
|
| 681 |
+
"fill": "ones",
|
| 682 |
+
"shape": [
|
| 683 |
+
1
|
| 684 |
+
],
|
| 685 |
+
"index": 37,
|
| 686 |
+
"input_name": "state_37",
|
| 687 |
+
"output_name": "out_state_37"
|
| 688 |
+
},
|
| 689 |
+
{
|
| 690 |
+
"module": "encoder.model.4.block.3",
|
| 691 |
+
"key": "previous",
|
| 692 |
+
"dtype": "float32",
|
| 693 |
+
"fill": "zeros",
|
| 694 |
+
"shape": [
|
| 695 |
+
1,
|
| 696 |
+
64,
|
| 697 |
+
0
|
| 698 |
+
],
|
| 699 |
+
"index": 38,
|
| 700 |
+
"input_name": "state_38",
|
| 701 |
+
"output_name": "out_state_38"
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"module": "encoder.model.6",
|
| 705 |
+
"key": "first",
|
| 706 |
+
"dtype": "bool",
|
| 707 |
+
"fill": "ones",
|
| 708 |
+
"shape": [
|
| 709 |
+
1
|
| 710 |
+
],
|
| 711 |
+
"index": 39,
|
| 712 |
+
"input_name": "state_39",
|
| 713 |
+
"output_name": "out_state_39"
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"module": "encoder.model.6",
|
| 717 |
+
"key": "previous",
|
| 718 |
+
"dtype": "float32",
|
| 719 |
+
"fill": "zeros",
|
| 720 |
+
"shape": [
|
| 721 |
+
1,
|
| 722 |
+
128,
|
| 723 |
+
5
|
| 724 |
+
],
|
| 725 |
+
"index": 40,
|
| 726 |
+
"input_name": "state_40",
|
| 727 |
+
"output_name": "out_state_40"
|
| 728 |
+
},
|
| 729 |
+
{
|
| 730 |
+
"module": "encoder.model.7.block.1",
|
| 731 |
+
"key": "first",
|
| 732 |
+
"dtype": "bool",
|
| 733 |
+
"fill": "ones",
|
| 734 |
+
"shape": [
|
| 735 |
+
1
|
| 736 |
+
],
|
| 737 |
+
"index": 41,
|
| 738 |
+
"input_name": "state_41",
|
| 739 |
+
"output_name": "out_state_41"
|
| 740 |
+
},
|
| 741 |
+
{
|
| 742 |
+
"module": "encoder.model.7.block.1",
|
| 743 |
+
"key": "previous",
|
| 744 |
+
"dtype": "float32",
|
| 745 |
+
"fill": "zeros",
|
| 746 |
+
"shape": [
|
| 747 |
+
1,
|
| 748 |
+
256,
|
| 749 |
+
2
|
| 750 |
+
],
|
| 751 |
+
"index": 42,
|
| 752 |
+
"input_name": "state_42",
|
| 753 |
+
"output_name": "out_state_42"
|
| 754 |
+
},
|
| 755 |
+
{
|
| 756 |
+
"module": "encoder.model.7.block.3",
|
| 757 |
+
"key": "first",
|
| 758 |
+
"dtype": "bool",
|
| 759 |
+
"fill": "ones",
|
| 760 |
+
"shape": [
|
| 761 |
+
1
|
| 762 |
+
],
|
| 763 |
+
"index": 43,
|
| 764 |
+
"input_name": "state_43",
|
| 765 |
+
"output_name": "out_state_43"
|
| 766 |
+
},
|
| 767 |
+
{
|
| 768 |
+
"module": "encoder.model.7.block.3",
|
| 769 |
+
"key": "previous",
|
| 770 |
+
"dtype": "float32",
|
| 771 |
+
"fill": "zeros",
|
| 772 |
+
"shape": [
|
| 773 |
+
1,
|
| 774 |
+
128,
|
| 775 |
+
0
|
| 776 |
+
],
|
| 777 |
+
"index": 44,
|
| 778 |
+
"input_name": "state_44",
|
| 779 |
+
"output_name": "out_state_44"
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"module": "encoder.model.9",
|
| 783 |
+
"key": "first",
|
| 784 |
+
"dtype": "bool",
|
| 785 |
+
"fill": "ones",
|
| 786 |
+
"shape": [
|
| 787 |
+
1
|
| 788 |
+
],
|
| 789 |
+
"index": 45,
|
| 790 |
+
"input_name": "state_45",
|
| 791 |
+
"output_name": "out_state_45"
|
| 792 |
+
},
|
| 793 |
+
{
|
| 794 |
+
"module": "encoder.model.9",
|
| 795 |
+
"key": "previous",
|
| 796 |
+
"dtype": "float32",
|
| 797 |
+
"fill": "zeros",
|
| 798 |
+
"shape": [
|
| 799 |
+
1,
|
| 800 |
+
256,
|
| 801 |
+
6
|
| 802 |
+
],
|
| 803 |
+
"index": 46,
|
| 804 |
+
"input_name": "state_46",
|
| 805 |
+
"output_name": "out_state_46"
|
| 806 |
+
},
|
| 807 |
+
{
|
| 808 |
+
"module": "encoder_transformer.transformer.layers.0.self_attn",
|
| 809 |
+
"key": "cache",
|
| 810 |
+
"dtype": "float32",
|
| 811 |
+
"fill": "nan",
|
| 812 |
+
"shape": [
|
| 813 |
+
2,
|
| 814 |
+
1,
|
| 815 |
+
8192,
|
| 816 |
+
8,
|
| 817 |
+
64
|
| 818 |
+
],
|
| 819 |
+
"index": 47,
|
| 820 |
+
"input_name": "state_47",
|
| 821 |
+
"output_name": "out_state_47"
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"module": "encoder_transformer.transformer.layers.0.self_attn",
|
| 825 |
+
"key": "offset",
|
| 826 |
+
"dtype": "int64",
|
| 827 |
+
"fill": "zeros",
|
| 828 |
+
"shape": [
|
| 829 |
+
1
|
| 830 |
+
],
|
| 831 |
+
"index": 48,
|
| 832 |
+
"input_name": "state_48",
|
| 833 |
+
"output_name": "out_state_48"
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"module": "encoder_transformer.transformer.layers.1.self_attn",
|
| 837 |
+
"key": "cache",
|
| 838 |
+
"dtype": "float32",
|
| 839 |
+
"fill": "nan",
|
| 840 |
+
"shape": [
|
| 841 |
+
2,
|
| 842 |
+
1,
|
| 843 |
+
8192,
|
| 844 |
+
8,
|
| 845 |
+
64
|
| 846 |
+
],
|
| 847 |
+
"index": 49,
|
| 848 |
+
"input_name": "state_49",
|
| 849 |
+
"output_name": "out_state_49"
|
| 850 |
+
},
|
| 851 |
+
{
|
| 852 |
+
"module": "encoder_transformer.transformer.layers.1.self_attn",
|
| 853 |
+
"key": "offset",
|
| 854 |
+
"dtype": "int64",
|
| 855 |
+
"fill": "zeros",
|
| 856 |
+
"shape": [
|
| 857 |
+
1
|
| 858 |
+
],
|
| 859 |
+
"index": 50,
|
| 860 |
+
"input_name": "state_50",
|
| 861 |
+
"output_name": "out_state_50"
|
| 862 |
+
},
|
| 863 |
+
{
|
| 864 |
+
"module": "upsample.convtr",
|
| 865 |
+
"key": "partial",
|
| 866 |
+
"dtype": "float32",
|
| 867 |
+
"fill": "zeros",
|
| 868 |
+
"shape": [
|
| 869 |
+
1,
|
| 870 |
+
512,
|
| 871 |
+
16
|
| 872 |
+
],
|
| 873 |
+
"index": 51,
|
| 874 |
+
"input_name": "state_51",
|
| 875 |
+
"output_name": "out_state_51"
|
| 876 |
+
}
|
| 877 |
+
],
|
| 878 |
+
"mimi_increment_per_frame": 16
|
| 879 |
+
}
|
pocket_ko/default_voice_p5.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14c9e068aa94f443e3d7bf32c98b45aa5581067beb26bf396a5673a24aa449e0
|
| 3 |
+
size 3147568
|
pocket_ko/flow_lm_flow_int8.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:268122a7f19c60e5735554474534282a22b0093a98529268a96e8080640f313f
|
| 3 |
+
size 9960222
|
pocket_ko/flow_lm_main_int8.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd779c7de439e8ab3c2eee55997b6821d5ee895fb2a8a31a55d2bda8feaed36b
|
| 3 |
+
size 76134646
|
pocket_ko/mimi_decoder.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b23fa3b5a0d4fc5552f50c25628e7a64a81be6564833bbc878e8b34e7f8ae708
|
| 3 |
+
size 41480010
|
pocket_ko/mimi_encoder.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a0d6a9ed6a1caa393a4af2b6a12231a2c25e782cc45688ac1ba43820364ff05
|
| 3 |
+
size 39250045
|
pocket_ko/model_manifest.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bundle.json": {
|
| 3 |
+
"size": 18586,
|
| 4 |
+
"sha256": "c347a9949e52cf56236609af52d922fa510b36466bf5f18f6ccba7ca1cee306d"
|
| 5 |
+
},
|
| 6 |
+
"tokenizer.model": {
|
| 7 |
+
"size": 63083,
|
| 8 |
+
"sha256": "930a09912fc1ef73f2cec8ab16ae3411cf4308b7e7ac7b477eb9a52d13034982"
|
| 9 |
+
},
|
| 10 |
+
"bos_before_voice.npy": {
|
| 11 |
+
"size": 4224,
|
| 12 |
+
"sha256": "e2400fb47b3f5d45ece082e5d563b4d8e87ee6d4b72782b8fbb432b2e5a6ae4d"
|
| 13 |
+
},
|
| 14 |
+
"text_conditioner.onnx": {
|
| 15 |
+
"size": 16388338,
|
| 16 |
+
"sha256": "0921775ff47831a93c5bcd10c58816b737262c0dbb50d50a1a80d4a50a878996"
|
| 17 |
+
},
|
| 18 |
+
"flow_lm_main_int8.onnx": {
|
| 19 |
+
"size": 76134646,
|
| 20 |
+
"sha256": "dd779c7de439e8ab3c2eee55997b6821d5ee895fb2a8a31a55d2bda8feaed36b"
|
| 21 |
+
},
|
| 22 |
+
"flow_lm_flow_int8.onnx": {
|
| 23 |
+
"size": 9960222,
|
| 24 |
+
"sha256": "268122a7f19c60e5735554474534282a22b0093a98529268a96e8080640f313f"
|
| 25 |
+
},
|
| 26 |
+
"mimi_decoder.onnx": {
|
| 27 |
+
"size": 41480010,
|
| 28 |
+
"sha256": "b23fa3b5a0d4fc5552f50c25628e7a64a81be6564833bbc878e8b34e7f8ae708"
|
| 29 |
+
},
|
| 30 |
+
"mimi_encoder.onnx": {
|
| 31 |
+
"size": 39250045,
|
| 32 |
+
"sha256": "1a0d6a9ed6a1caa393a4af2b6a12231a2c25e782cc45688ac1ba43820364ff05"
|
| 33 |
+
},
|
| 34 |
+
"default_voice_p5.safetensors": {
|
| 35 |
+
"size": 3147568,
|
| 36 |
+
"sha256": "14c9e068aa94f443e3d7bf32c98b45aa5581067beb26bf396a5673a24aa449e0"
|
| 37 |
+
}
|
| 38 |
+
}
|
pocket_ko/text_conditioner.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0921775ff47831a93c5bcd10c58816b737262c0dbb50d50a1a80d4a50a878996
|
| 3 |
+
size 16388338
|
pocket_ko/tokenizer.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:930a09912fc1ef73f2cec8ab16ae3411cf4308b7e7ac7b477eb9a52d13034982
|
| 3 |
+
size 63083
|