Upload config.json
Browse files- config.json +41 -0
config.json
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_postnet": true,
|
| 3 |
+
"architectures": [
|
| 4 |
+
"FastSpeech2Model"
|
| 5 |
+
],
|
| 6 |
+
"attention_dropout": 0,
|
| 7 |
+
"decoder_attention_heads": 2,
|
| 8 |
+
"decoder_embed_dim": 256,
|
| 9 |
+
"decoder_layers": 4,
|
| 10 |
+
"dropout": 0.2,
|
| 11 |
+
"encoder_attention_heads": 2,
|
| 12 |
+
"encoder_embed_dim": 256,
|
| 13 |
+
"encoder_layers": 4,
|
| 14 |
+
"energy_max": 5.646703243255615,
|
| 15 |
+
"energy_min": -3.548804759979248,
|
| 16 |
+
"fft_hidden_dim": 1024,
|
| 17 |
+
"fft_kernel_size": 9,
|
| 18 |
+
"initializer_range": 0.0625,
|
| 19 |
+
"max_source_positions": 1024,
|
| 20 |
+
"mean": true,
|
| 21 |
+
"model_type": "fastspeech2",
|
| 22 |
+
"n_frames_per_step": 1,
|
| 23 |
+
"num_speakers": 200,
|
| 24 |
+
"output_frame_dim": 80,
|
| 25 |
+
"pad_token_id": 1,
|
| 26 |
+
"pitch_max": -2.5049682966705484,
|
| 27 |
+
"pitch_min": -2.5049682966705484,
|
| 28 |
+
"postnet_conv_dim": 512,
|
| 29 |
+
"postnet_conv_kernel_size": 5,
|
| 30 |
+
"postnet_dropout": 0.5,
|
| 31 |
+
"postnet_layers": 5,
|
| 32 |
+
"speaker_embed_dim": 64,
|
| 33 |
+
"std": true,
|
| 34 |
+
"torch_dtype": "float32",
|
| 35 |
+
"transformers_version": "4.17.0.dev0",
|
| 36 |
+
"var_pred_dropout": 0.5,
|
| 37 |
+
"var_pred_hidden_dim": 256,
|
| 38 |
+
"var_pred_kernel_size": 3,
|
| 39 |
+
"var_pred_n_bins": 256,
|
| 40 |
+
"vocab_size": 74
|
| 41 |
+
}
|