Upload 2 files
Browse files- config.json +68 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,68 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "pocket_tts",
|
| 3 |
+
"flow_lm": {
|
| 4 |
+
"dtype": "float32",
|
| 5 |
+
"flow": {
|
| 6 |
+
"dim": 512,
|
| 7 |
+
"depth": 6
|
| 8 |
+
},
|
| 9 |
+
"transformer": {
|
| 10 |
+
"hidden_scale": 4,
|
| 11 |
+
"max_period": 10000,
|
| 12 |
+
"d_model": 1024,
|
| 13 |
+
"num_heads": 16,
|
| 14 |
+
"num_layers": 6
|
| 15 |
+
},
|
| 16 |
+
"lookup_table": {
|
| 17 |
+
"dim": 1024,
|
| 18 |
+
"n_bins": 4000,
|
| 19 |
+
"tokenizer": "sentencepiece",
|
| 20 |
+
"tokenizer_path": "hf://kyutai/pocket-tts-without-voice-cloning/tokenizer.model@d4fdd22ae8c8e1cb3634e150ebeff1dab2d16df3"
|
| 21 |
+
},
|
| 22 |
+
"weights_path": null
|
| 23 |
+
},
|
| 24 |
+
"mimi": {
|
| 25 |
+
"dtype": "float32",
|
| 26 |
+
"sample_rate": 24000,
|
| 27 |
+
"channels": 1,
|
| 28 |
+
"frame_rate": 12.5,
|
| 29 |
+
"seanet": {
|
| 30 |
+
"dimension": 512,
|
| 31 |
+
"channels": 1,
|
| 32 |
+
"n_filters": 64,
|
| 33 |
+
"n_residual_layers": 1,
|
| 34 |
+
"ratios": [
|
| 35 |
+
6,
|
| 36 |
+
5,
|
| 37 |
+
4
|
| 38 |
+
],
|
| 39 |
+
"kernel_size": 7,
|
| 40 |
+
"residual_kernel_size": 3,
|
| 41 |
+
"last_kernel_size": 3,
|
| 42 |
+
"dilation_base": 2,
|
| 43 |
+
"pad_mode": "constant",
|
| 44 |
+
"compress": 2
|
| 45 |
+
},
|
| 46 |
+
"transformer": {
|
| 47 |
+
"d_model": 512,
|
| 48 |
+
"input_dimension": 512,
|
| 49 |
+
"output_dimensions": [
|
| 50 |
+
512
|
| 51 |
+
],
|
| 52 |
+
"num_heads": 8,
|
| 53 |
+
"num_layers": 2,
|
| 54 |
+
"layer_scale": 0.01,
|
| 55 |
+
"context": 250,
|
| 56 |
+
"dim_feedforward": 2048,
|
| 57 |
+
"max_period": 10000.0
|
| 58 |
+
},
|
| 59 |
+
"quantizer": {
|
| 60 |
+
"dimension": 32,
|
| 61 |
+
"output_dimension": 512
|
| 62 |
+
},
|
| 63 |
+
"weights_path": null
|
| 64 |
+
},
|
| 65 |
+
"weights_path": "hf://kyutai/pocket-tts/tts_b6369a24.safetensors@427e3d61b276ed69fdd03de0d185fa8a8d97fc5b",
|
| 66 |
+
"weights_path_without_voice_cloning": "hf://kyutai/pocket-tts-without-voice-cloning/tts_b6369a24.safetensors@d4fdd22ae8c8e1cb3634e150ebeff1dab2d16df3",
|
| 67 |
+
"model_path": null
|
| 68 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:60ddd85019dddbe6c1d220e311ca5fc753972978a89f563c0bbe1ae943120072
|
| 3 |
+
size 235739497
|