Initial commit
Browse files- best_model.pth +3 -0
- best_model_75582.pth +3 -0
- checkpoint_30000.pth +3 -0
- checkpoint_40000.pth +3 -0
- checkpoint_50000.pth +3 -0
- checkpoint_60000.pth +3 -0
- checkpoint_70000.pth +3 -0
- config.json +183 -0
- events.out.tfevents.1651047468.dcf8fa084f7d +3 -0
- trainer_0_log.txt +0 -0
best_model.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:093cbb6fb612ac5db27a5c667fded16d90918d884f0f6fbd923b48cd20dc3e2f
|
3 |
+
size 343669287
|
best_model_75582.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:093cbb6fb612ac5db27a5c667fded16d90918d884f0f6fbd923b48cd20dc3e2f
|
3 |
+
size 343669287
|
checkpoint_30000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:48e2fe3ae8dbf5369641d45784897bd4ff75da04de82630ddae4cec40856798f
|
3 |
+
size 343668263
|
checkpoint_40000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8c3f44e250597f677e115456e19d4094f95b7be75c9fed2c55f14122587ad8f5
|
3 |
+
size 343668263
|
checkpoint_50000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2c1ed9391174d4e1afad0ebdaa79d05c3e6ed2c55048ccdb36acf10f89edf125
|
3 |
+
size 343668263
|
checkpoint_60000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b76adc19fa7f97a8a8bf3919786cb3ed6a6c09892788ee6407f07442e7031b26
|
3 |
+
size 343668263
|
checkpoint_70000.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8b4ca5616838794aa94dce676f9daab114a692756f7bb10027251f8c4593050b
|
3 |
+
size 343669287
|
config.json
ADDED
@@ -0,0 +1,183 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"output_path": "/root/TTS",
|
3 |
+
"logger_uri": null,
|
4 |
+
"run_name": "run",
|
5 |
+
"project_name": null,
|
6 |
+
"run_description": "\ud83d\udc38Coqui trainer run.",
|
7 |
+
"print_step": 5,
|
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": 1000,
|
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_deterministic": false,
|
46 |
+
"cudnn_benchmark": true,
|
47 |
+
"training_seed": 54321,
|
48 |
+
"model": "glow_tts",
|
49 |
+
"num_loader_workers": 8,
|
50 |
+
"num_eval_loader_workers": 4,
|
51 |
+
"use_noise_augment": 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": 22050,
|
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": null,
|
89 |
+
"compute_input_seq_cache": false,
|
90 |
+
"text_cleaner": "basic_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": "<PAD>",
|
98 |
+
"eos": "<EOS>",
|
99 |
+
"bos": "<BOS>",
|
100 |
+
"blank": "<BLNK>",
|
101 |
+
"characters": "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz",
|
102 |
+
"punctuations": "!'(),-.:;? ",
|
103 |
+
"phonemes": null,
|
104 |
+
"is_unique": false,
|
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": "/root/TTS/./AudiobookSpeech/",
|
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 |
+
"It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.",
|
132 |
+
"Be a voice, not an echo.",
|
133 |
+
"I'm sorry Dave. I'm afraid I can't do that.",
|
134 |
+
"This cake is great. It's so delicious and moist.",
|
135 |
+
"Prior to November 22, 1963."
|
136 |
+
],
|
137 |
+
"eval_split_max_size": null,
|
138 |
+
"eval_split_size": 0.01,
|
139 |
+
"use_speaker_weighted_sampler": false,
|
140 |
+
"speaker_weighted_sampler_alpha": 1.0,
|
141 |
+
"use_language_weighted_sampler": false,
|
142 |
+
"language_weighted_sampler_alpha": 1.0,
|
143 |
+
"num_chars": 67,
|
144 |
+
"encoder_type": "rel_pos_transformer",
|
145 |
+
"encoder_params": {
|
146 |
+
"kernel_size": 3,
|
147 |
+
"dropout_p": 0.1,
|
148 |
+
"num_layers": 6,
|
149 |
+
"num_heads": 2,
|
150 |
+
"hidden_channels_ffn": 768,
|
151 |
+
"input_length": null
|
152 |
+
},
|
153 |
+
"use_encoder_prenet": true,
|
154 |
+
"hidden_channels_enc": 192,
|
155 |
+
"hidden_channels_dec": 192,
|
156 |
+
"hidden_channels_dp": 256,
|
157 |
+
"dropout_p_dp": 0.1,
|
158 |
+
"dropout_p_dec": 0.05,
|
159 |
+
"mean_only": true,
|
160 |
+
"out_channels": 80,
|
161 |
+
"num_flow_blocks_dec": 12,
|
162 |
+
"inference_noise_scale": 0.0,
|
163 |
+
"kernel_size_dec": 5,
|
164 |
+
"dilation_rate": 1,
|
165 |
+
"num_block_layers": 4,
|
166 |
+
"num_speakers": 0,
|
167 |
+
"c_in_channels": 0,
|
168 |
+
"num_splits": 4,
|
169 |
+
"num_squeeze": 2,
|
170 |
+
"sigmoid_scale": false,
|
171 |
+
"d_vector_dim": 0,
|
172 |
+
"data_dep_init_steps": 10,
|
173 |
+
"style_wav_for_test": null,
|
174 |
+
"length_scale": 1.0,
|
175 |
+
"use_speaker_embedding": false,
|
176 |
+
"speakers_file": null,
|
177 |
+
"use_d_vector_file": false,
|
178 |
+
"d_vector_file": false,
|
179 |
+
"min_seq_len": 3,
|
180 |
+
"max_seq_len": 500,
|
181 |
+
"r": 1,
|
182 |
+
"github_branch": "* main"
|
183 |
+
}
|
events.out.tfevents.1651047468.dcf8fa084f7d
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:48744c813f7117be378c07a2a5fd96b3959b956cdbb4aba38ba88839b044cd0a
|
3 |
+
size 543235614
|
trainer_0_log.txt
ADDED
The diff for this file is too large to render.
See raw diff
|
|