Upload 2 files
Browse files- eng_lad_v2_checkpoint_770000.pth +3 -0
- eng_lad_v2_config.json +175 -0
eng_lad_v2_checkpoint_770000.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7ab8a393a61f2f79311166eb36b4a51deebbb1a3f9257c99e71f1fcb724b866
|
| 3 |
+
size 343694311
|
eng_lad_v2_config.json
ADDED
|
@@ -0,0 +1,175 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"output_path": "/homedtcl/rzevallos/TTS/TTS/recipes/ladino",
|
| 3 |
+
"logger_uri": null,
|
| 4 |
+
"run_name": "Eng_Lad",
|
| 5 |
+
"project_name": null,
|
| 6 |
+
"run_description": "English_Ladino_test",
|
| 7 |
+
"print_step": 25,
|
| 8 |
+
"plot_step": 100,
|
| 9 |
+
"model_param_stats": false,
|
| 10 |
+
"wandb_entity": null,
|
| 11 |
+
"dashboard_logger": "tensorboard",
|
| 12 |
+
"log_model_step": null,
|
| 13 |
+
"save_step": 10000,
|
| 14 |
+
"save_n_checkpoints": 5,
|
| 15 |
+
"save_checkpoints": true,
|
| 16 |
+
"save_all_best": false,
|
| 17 |
+
"save_best_after": 10000,
|
| 18 |
+
"target_loss": null,
|
| 19 |
+
"print_eval": false,
|
| 20 |
+
"test_delay_epochs": -1,
|
| 21 |
+
"run_eval": true,
|
| 22 |
+
"distributed_backend": "nccl",
|
| 23 |
+
"distributed_url": "tcp://localhost:54321",
|
| 24 |
+
"mixed_precision": true,
|
| 25 |
+
"epochs": 5000,
|
| 26 |
+
"batch_size": 32,
|
| 27 |
+
"eval_batch_size": 16,
|
| 28 |
+
"grad_clip": 5.0,
|
| 29 |
+
"scheduler_after_epoch": true,
|
| 30 |
+
"lr": 0.001,
|
| 31 |
+
"optimizer": "RAdam",
|
| 32 |
+
"optimizer_params": {
|
| 33 |
+
"betas": [
|
| 34 |
+
0.9,
|
| 35 |
+
0.998
|
| 36 |
+
],
|
| 37 |
+
"weight_decay": 1e-06
|
| 38 |
+
},
|
| 39 |
+
"lr_scheduler": "NoamLR",
|
| 40 |
+
"lr_scheduler_params": {
|
| 41 |
+
"warmup_steps": 4000
|
| 42 |
+
},
|
| 43 |
+
"use_grad_scaler": false,
|
| 44 |
+
"cudnn_enable": true,
|
| 45 |
+
"cudnn_benchmark": true,
|
| 46 |
+
"torch_seed": 54321,
|
| 47 |
+
"model": "glow_tts",
|
| 48 |
+
"num_loader_workers": 4,
|
| 49 |
+
"num_eval_loader_workers": 4,
|
| 50 |
+
"use_noise_augment": false,
|
| 51 |
+
"use_language_weighted_sampler": false,
|
| 52 |
+
"audio": {
|
| 53 |
+
"fft_size": 1024,
|
| 54 |
+
"win_length": 1024,
|
| 55 |
+
"hop_length": 256,
|
| 56 |
+
"frame_shift_ms": null,
|
| 57 |
+
"frame_length_ms": null,
|
| 58 |
+
"stft_pad_mode": "reflect",
|
| 59 |
+
"sample_rate": 16000,
|
| 60 |
+
"resample": false,
|
| 61 |
+
"preemphasis": 0.0,
|
| 62 |
+
"ref_level_db": 20,
|
| 63 |
+
"do_sound_norm": false,
|
| 64 |
+
"log_func": "np.log10",
|
| 65 |
+
"do_trim_silence": true,
|
| 66 |
+
"trim_db": 45,
|
| 67 |
+
"do_rms_norm": false,
|
| 68 |
+
"db_level": null,
|
| 69 |
+
"power": 1.5,
|
| 70 |
+
"griffin_lim_iters": 60,
|
| 71 |
+
"num_mels": 80,
|
| 72 |
+
"mel_fmin": 0.0,
|
| 73 |
+
"mel_fmax": null,
|
| 74 |
+
"spec_gain": 20,
|
| 75 |
+
"do_amp_to_db_linear": true,
|
| 76 |
+
"do_amp_to_db_mel": true,
|
| 77 |
+
"pitch_fmax": 640.0,
|
| 78 |
+
"pitch_fmin": 0.0,
|
| 79 |
+
"signal_norm": true,
|
| 80 |
+
"min_level_db": -100,
|
| 81 |
+
"symmetric_norm": true,
|
| 82 |
+
"max_norm": 4.0,
|
| 83 |
+
"clip_norm": true,
|
| 84 |
+
"stats_path": null
|
| 85 |
+
},
|
| 86 |
+
"use_phonemes": false,
|
| 87 |
+
"phonemizer": null,
|
| 88 |
+
"phoneme_language": "en-us",
|
| 89 |
+
"compute_input_seq_cache": false,
|
| 90 |
+
"text_cleaner": "multilingual_cleaners",
|
| 91 |
+
"enable_eos_bos_chars": false,
|
| 92 |
+
"test_sentences_file": "",
|
| 93 |
+
"phoneme_cache_path": null,
|
| 94 |
+
"characters": {
|
| 95 |
+
"characters_class": "TTS.tts.utils.text.characters.Graphemes",
|
| 96 |
+
"vocab_dict": null,
|
| 97 |
+
"pad": "_",
|
| 98 |
+
"eos": "*",
|
| 99 |
+
"bos": "^",
|
| 100 |
+
"blank": null,
|
| 101 |
+
"characters": "ABCDEFGHIJKLMNOPRSTUVXYZabcdefghijklmnopqrstuvwxyz\u00c7\u00d6\u00dc\u00e2\u00e7\u00e8\u00f6\u00fa\u00fc\u011f\u0130\u0131\u015e\u015f",
|
| 102 |
+
"punctuations": "!'(),-.:;? ",
|
| 103 |
+
"phonemes": "",
|
| 104 |
+
"is_unique": true,
|
| 105 |
+
"is_sorted": true
|
| 106 |
+
},
|
| 107 |
+
"add_blank": false,
|
| 108 |
+
"batch_group_size": 0,
|
| 109 |
+
"loss_masking": null,
|
| 110 |
+
"sort_by_audio_len": false,
|
| 111 |
+
"min_audio_len": 1,
|
| 112 |
+
"max_audio_len": Infinity,
|
| 113 |
+
"min_text_len": 1,
|
| 114 |
+
"max_text_len": Infinity,
|
| 115 |
+
"compute_f0": false,
|
| 116 |
+
"compute_linear_spec": false,
|
| 117 |
+
"precompute_num_workers": 0,
|
| 118 |
+
"start_by_longest": false,
|
| 119 |
+
"datasets": [
|
| 120 |
+
{
|
| 121 |
+
"name": "ljspeech",
|
| 122 |
+
"path": "/homedtcl/rzevallos/TTS/recipes/ljspeech/LJSpeech-1.1",
|
| 123 |
+
"meta_file_train": "metadata.csv",
|
| 124 |
+
"ignored_speakers": null,
|
| 125 |
+
"language": "",
|
| 126 |
+
"meta_file_val": "",
|
| 127 |
+
"meta_file_attn_mask": ""
|
| 128 |
+
}
|
| 129 |
+
],
|
| 130 |
+
"test_sentences": [
|
| 131 |
+
"Hello my name is Rodolfo",
|
| 132 |
+
"Hello my name is Eva"
|
| 133 |
+
],
|
| 134 |
+
"eval_split_max_size": null,
|
| 135 |
+
"eval_split_size": 0.01,
|
| 136 |
+
"num_chars": 78,
|
| 137 |
+
"encoder_type": "rel_pos_transformer",
|
| 138 |
+
"encoder_params": {
|
| 139 |
+
"kernel_size": 3,
|
| 140 |
+
"dropout_p": 0.1,
|
| 141 |
+
"num_layers": 6,
|
| 142 |
+
"num_heads": 2,
|
| 143 |
+
"hidden_channels_ffn": 768,
|
| 144 |
+
"input_length": null
|
| 145 |
+
},
|
| 146 |
+
"use_encoder_prenet": true,
|
| 147 |
+
"hidden_channels_enc": 192,
|
| 148 |
+
"hidden_channels_dec": 192,
|
| 149 |
+
"hidden_channels_dp": 256,
|
| 150 |
+
"dropout_p_dp": 0.1,
|
| 151 |
+
"dropout_p_dec": 0.05,
|
| 152 |
+
"mean_only": true,
|
| 153 |
+
"out_channels": 80,
|
| 154 |
+
"num_flow_blocks_dec": 12,
|
| 155 |
+
"inference_noise_scale": 0.0,
|
| 156 |
+
"kernel_size_dec": 5,
|
| 157 |
+
"dilation_rate": 1,
|
| 158 |
+
"num_block_layers": 4,
|
| 159 |
+
"num_speakers": 0,
|
| 160 |
+
"c_in_channels": 0,
|
| 161 |
+
"num_splits": 4,
|
| 162 |
+
"num_squeeze": 2,
|
| 163 |
+
"sigmoid_scale": false,
|
| 164 |
+
"d_vector_dim": 0,
|
| 165 |
+
"data_dep_init_steps": 10,
|
| 166 |
+
"style_wav_for_test": null,
|
| 167 |
+
"length_scale": 1.0,
|
| 168 |
+
"use_speaker_embedding": false,
|
| 169 |
+
"speakers_file": null,
|
| 170 |
+
"use_d_vector_file": false,
|
| 171 |
+
"d_vector_file": false,
|
| 172 |
+
"min_seq_len": 3,
|
| 173 |
+
"max_seq_len": 500,
|
| 174 |
+
"r": 1
|
| 175 |
+
}
|