Upload config.json with huggingface_hub
Browse files- config.json +117 -0
config.json
ADDED
|
@@ -0,0 +1,117 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"backbone": {
|
| 3 |
+
"d_model": 2048,
|
| 4 |
+
"d_intermediate": 0,
|
| 5 |
+
"attn_mlp_d_intermediate": 8192,
|
| 6 |
+
"n_layer": 46,
|
| 7 |
+
"ssm_cfg": {
|
| 8 |
+
"layer": "Mamba2"
|
| 9 |
+
},
|
| 10 |
+
"attn_layer_idx": [
|
| 11 |
+
0,
|
| 12 |
+
4,
|
| 13 |
+
8,
|
| 14 |
+
12,
|
| 15 |
+
16,
|
| 16 |
+
20,
|
| 17 |
+
24,
|
| 18 |
+
28,
|
| 19 |
+
32,
|
| 20 |
+
36,
|
| 21 |
+
40,
|
| 22 |
+
44
|
| 23 |
+
],
|
| 24 |
+
"attn_cfg": {
|
| 25 |
+
"causal": true,
|
| 26 |
+
"num_heads": 16,
|
| 27 |
+
"num_heads_kv": 4,
|
| 28 |
+
"rotary_emb_dim": 128,
|
| 29 |
+
"qkv_proj_bias": false,
|
| 30 |
+
"out_proj_bias": false
|
| 31 |
+
},
|
| 32 |
+
"rms_norm": false,
|
| 33 |
+
"residual_in_fp32": false,
|
| 34 |
+
"norm_epsilon": 1e-05
|
| 35 |
+
},
|
| 36 |
+
"prefix_conditioner": {
|
| 37 |
+
"conditioners": [
|
| 38 |
+
{
|
| 39 |
+
"type": "EspeakPhonemeConditioner",
|
| 40 |
+
"name": "espeak"
|
| 41 |
+
},
|
| 42 |
+
{
|
| 43 |
+
"cond_dim": 128,
|
| 44 |
+
"uncond_type": "learned",
|
| 45 |
+
"projection": "linear",
|
| 46 |
+
"type": "PassthroughConditioner",
|
| 47 |
+
"name": "speaker"
|
| 48 |
+
},
|
| 49 |
+
{
|
| 50 |
+
"input_dim": 8,
|
| 51 |
+
"uncond_type": "learned",
|
| 52 |
+
"type": "FourierConditioner",
|
| 53 |
+
"name": "emotion"
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"min_val": 0,
|
| 57 |
+
"max_val": 24000,
|
| 58 |
+
"uncond_type": "learned",
|
| 59 |
+
"type": "FourierConditioner",
|
| 60 |
+
"name": "fmax"
|
| 61 |
+
},
|
| 62 |
+
{
|
| 63 |
+
"min_val": 0,
|
| 64 |
+
"max_val": 400,
|
| 65 |
+
"uncond_type": "learned",
|
| 66 |
+
"type": "FourierConditioner",
|
| 67 |
+
"name": "pitch_std"
|
| 68 |
+
},
|
| 69 |
+
{
|
| 70 |
+
"min_val": 0,
|
| 71 |
+
"max_val": 40,
|
| 72 |
+
"uncond_type": "learned",
|
| 73 |
+
"type": "FourierConditioner",
|
| 74 |
+
"name": "speaking_rate"
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"min_val": -1,
|
| 78 |
+
"max_val": 126,
|
| 79 |
+
"uncond_type": "learned",
|
| 80 |
+
"type": "IntegerConditioner",
|
| 81 |
+
"name": "language_id"
|
| 82 |
+
},
|
| 83 |
+
{
|
| 84 |
+
"input_dim": 8,
|
| 85 |
+
"min_val": 0.5,
|
| 86 |
+
"max_val": 0.8,
|
| 87 |
+
"uncond_type": "learned",
|
| 88 |
+
"type": "FourierConditioner",
|
| 89 |
+
"name": "vqscore_8"
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"min_val": -1.0,
|
| 93 |
+
"max_val": 1000,
|
| 94 |
+
"uncond_type": "learned",
|
| 95 |
+
"type": "FourierConditioner",
|
| 96 |
+
"name": "ctc_loss"
|
| 97 |
+
},
|
| 98 |
+
{
|
| 99 |
+
"min_val": 1,
|
| 100 |
+
"max_val": 5,
|
| 101 |
+
"uncond_type": "learned",
|
| 102 |
+
"type": "FourierConditioner",
|
| 103 |
+
"name": "dnsmos_ovrl"
|
| 104 |
+
},
|
| 105 |
+
{
|
| 106 |
+
"min_val": 0,
|
| 107 |
+
"max_val": 1,
|
| 108 |
+
"uncond_type": "learned",
|
| 109 |
+
"type": "IntegerConditioner",
|
| 110 |
+
"name": "speaker_noised"
|
| 111 |
+
}
|
| 112 |
+
],
|
| 113 |
+
"projection": "linear"
|
| 114 |
+
},
|
| 115 |
+
"eos_token_id": 1024,
|
| 116 |
+
"masked_token_id": 1025
|
| 117 |
+
}
|