Upload config.json
Browse files- config.json +81 -0
config.json
ADDED
|
@@ -0,0 +1,81 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "xtts",
|
| 3 |
+
"audio": {
|
| 4 |
+
"sample_rate": 22050,
|
| 5 |
+
"output_sample_rate": 24000,
|
| 6 |
+
"dvae_sample_rate": 22050
|
| 7 |
+
},
|
| 8 |
+
"model_args": {
|
| 9 |
+
"gpt_batch_size": 1,
|
| 10 |
+
"enable_redaction": false,
|
| 11 |
+
"kv_cache": true,
|
| 12 |
+
"gpt_checkpoint": "model.pth",
|
| 13 |
+
"clvp_checkpoint": null,
|
| 14 |
+
"decoder_checkpoint": null,
|
| 15 |
+
"num_chars": 255,
|
| 16 |
+
"tokenizer_file": "vocab.json",
|
| 17 |
+
"gpt_max_audio_tokens": 605,
|
| 18 |
+
"gpt_max_text_tokens": 402,
|
| 19 |
+
"gpt_max_prompt_tokens": 70,
|
| 20 |
+
"gpt_layers": 30,
|
| 21 |
+
"gpt_n_model_channels": 1024,
|
| 22 |
+
"gpt_n_heads": 16,
|
| 23 |
+
"gpt_number_text_tokens": 8456,
|
| 24 |
+
"gpt_start_text_token": 261,
|
| 25 |
+
"gpt_stop_text_token": 0,
|
| 26 |
+
"gpt_num_audio_tokens": 1026,
|
| 27 |
+
"gpt_start_audio_token": 1024,
|
| 28 |
+
"gpt_stop_audio_token": 1025,
|
| 29 |
+
"gpt_code_stride_len": 1024,
|
| 30 |
+
"gpt_use_masking_gt_prompt_approach": true,
|
| 31 |
+
"gpt_use_perceiver_resampler": true,
|
| 32 |
+
"input_sample_rate": 22050,
|
| 33 |
+
"output_sample_rate": 24000,
|
| 34 |
+
"output_hop_length": 256,
|
| 35 |
+
"decoder_input_dim": 1024,
|
| 36 |
+
"d_vector_dim": 512,
|
| 37 |
+
"cond_d_vector_in_each_upsampling_layer": true,
|
| 38 |
+
"duration_const": 102400,
|
| 39 |
+
"min_conditioning_length": 11025,
|
| 40 |
+
"max_conditioning_length": 132300,
|
| 41 |
+
"gpt_loss_text_ce_weight": 0.01,
|
| 42 |
+
"gpt_loss_mel_ce_weight": 1.0,
|
| 43 |
+
"debug_loading_failures": false,
|
| 44 |
+
"max_wav_length": 330750,
|
| 45 |
+
"max_text_length": 400,
|
| 46 |
+
"mel_norm_file": "mel_stats.pth",
|
| 47 |
+
"dvae_checkpoint": "dvae.pth",
|
| 48 |
+
"xtts_checkpoint": "model.pth",
|
| 49 |
+
"vocoder": ""
|
| 50 |
+
},
|
| 51 |
+
"languages": [
|
| 52 |
+
"en",
|
| 53 |
+
"es",
|
| 54 |
+
"fr",
|
| 55 |
+
"de",
|
| 56 |
+
"it",
|
| 57 |
+
"pt",
|
| 58 |
+
"pl",
|
| 59 |
+
"tr",
|
| 60 |
+
"ru",
|
| 61 |
+
"nl",
|
| 62 |
+
"cs",
|
| 63 |
+
"ar",
|
| 64 |
+
"zh-cn",
|
| 65 |
+
"hu",
|
| 66 |
+
"ko",
|
| 67 |
+
"ja",
|
| 68 |
+
"hi",
|
| 69 |
+
"mns"
|
| 70 |
+
],
|
| 71 |
+
"temperature": 0.75,
|
| 72 |
+
"length_penalty": 1.0,
|
| 73 |
+
"repetition_penalty": 5.0,
|
| 74 |
+
"top_k": 50,
|
| 75 |
+
"top_p": 0.85,
|
| 76 |
+
"num_gpt_outputs": 1,
|
| 77 |
+
"gpt_cond_len": 30,
|
| 78 |
+
"gpt_cond_chunk_len": 4,
|
| 79 |
+
"max_ref_len": 30,
|
| 80 |
+
"sound_norm_refs": false
|
| 81 |
+
}
|