Actualización modelo
Browse files- sabela.pth +2 -2
- sabela_config.json +42 -26
sabela.pth
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:24714543c0735c1a2c4e7c79462ba3ec834f8d4afc28cbaa23bea3ee8f67964c
|
3 |
+
size 997811318
|
sabela_config.json
CHANGED
@@ -1,7 +1,7 @@
|
|
1 |
{
|
2 |
-
"output_path": "/home/
|
3 |
"logger_uri": null,
|
4 |
-
"run_name": "
|
5 |
"project_name": null,
|
6 |
"run_description": "\ud83d\udc38Coqui trainer run.",
|
7 |
"print_step": 25,
|
@@ -9,12 +9,13 @@
|
|
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":
|
15 |
"save_checkpoints": true,
|
16 |
"save_all_best": false,
|
17 |
-
"save_best_after":
|
18 |
"target_loss": null,
|
19 |
"print_eval": true,
|
20 |
"test_delay_epochs": -1,
|
@@ -23,15 +24,16 @@
|
|
23 |
"distributed_backend": "nccl",
|
24 |
"distributed_url": "tcp://localhost:54321",
|
25 |
"mixed_precision": true,
|
26 |
-
"
|
27 |
-
"
|
|
|
28 |
"eval_batch_size": 16,
|
29 |
"grad_clip": [
|
30 |
1000,
|
31 |
1000
|
32 |
],
|
33 |
"scheduler_after_epoch": true,
|
34 |
-
"lr": 0.
|
35 |
"optimizer": "AdamW",
|
36 |
"optimizer_params": {
|
37 |
"betas": [
|
@@ -41,9 +43,10 @@
|
|
41 |
"eps": 1e-09,
|
42 |
"weight_decay": 0.01
|
43 |
},
|
44 |
-
"lr_scheduler":
|
45 |
"lr_scheduler_params": {},
|
46 |
"use_grad_scaler": false,
|
|
|
47 |
"cudnn_enable": true,
|
48 |
"cudnn_deterministic": false,
|
49 |
"cudnn_benchmark": false,
|
@@ -63,22 +66,22 @@
|
|
63 |
},
|
64 |
"use_phonemes": false,
|
65 |
"phonemizer": null,
|
66 |
-
"phoneme_language":
|
67 |
"compute_input_seq_cache": true,
|
68 |
-
"text_cleaner":
|
69 |
"enable_eos_bos_chars": false,
|
70 |
"test_sentences_file": "",
|
71 |
-
"phoneme_cache_path": "/home/
|
72 |
"characters": {
|
73 |
"characters_class": "TTS.tts.models.vits.VitsCharacters",
|
74 |
"vocab_dict": null,
|
75 |
-
"pad": "
|
76 |
-
"eos": "
|
77 |
-
"bos": "
|
78 |
-
"blank":
|
79 |
-
"characters": "
|
80 |
-
"punctuations": "!\
|
81 |
-
"phonemes":
|
82 |
"is_unique": true,
|
83 |
"is_sorted": true
|
84 |
},
|
@@ -90,26 +93,39 @@
|
|
90 |
"min_text_len": 1,
|
91 |
"max_text_len": Infinity,
|
92 |
"compute_f0": false,
|
|
|
93 |
"compute_linear_spec": true,
|
94 |
"precompute_num_workers": 0,
|
95 |
"start_by_longest": false,
|
|
|
|
|
96 |
"datasets": [
|
97 |
{
|
98 |
-
"
|
99 |
-
"
|
100 |
-
"
|
|
|
101 |
"ignored_speakers": null,
|
102 |
"language": "",
|
|
|
103 |
"meta_file_val": "",
|
104 |
"meta_file_attn_mask": ""
|
105 |
}
|
106 |
],
|
107 |
"test_sentences": [
|
108 |
-
"
|
109 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
],
|
111 |
"eval_split_max_size": null,
|
112 |
-
"eval_split_size": 0.
|
113 |
"use_speaker_weighted_sampler": false,
|
114 |
"speaker_weighted_sampler_alpha": 1.0,
|
115 |
"use_language_weighted_sampler": false,
|
@@ -117,7 +133,7 @@
|
|
117 |
"use_length_weighted_sampler": false,
|
118 |
"length_weighted_sampler_alpha": 1.0,
|
119 |
"model_args": {
|
120 |
-
"num_chars":
|
121 |
"out_channels": 513,
|
122 |
"spec_segment_size": 32,
|
123 |
"hidden_channels": 192,
|
@@ -244,5 +260,5 @@
|
|
244 |
"use_d_vector_file": false,
|
245 |
"d_vector_file": null,
|
246 |
"d_vector_dim": 0,
|
247 |
-
"github_branch": "
|
248 |
}
|
|
|
1 |
{
|
2 |
+
"output_path": "/mnt/netapp2/Store_uni/home/usc/ci/ams/experimentos/sabela/normal",
|
3 |
"logger_uri": null,
|
4 |
+
"run_name": "sabela_48_104",
|
5 |
"project_name": null,
|
6 |
"run_description": "\ud83d\udc38Coqui trainer run.",
|
7 |
"print_step": 25,
|
|
|
9 |
"model_param_stats": false,
|
10 |
"wandb_entity": null,
|
11 |
"dashboard_logger": "tensorboard",
|
12 |
+
"save_on_interrupt": true,
|
13 |
"log_model_step": null,
|
14 |
"save_step": 10000,
|
15 |
+
"save_n_checkpoints": 10,
|
16 |
"save_checkpoints": true,
|
17 |
"save_all_best": false,
|
18 |
+
"save_best_after": 400,
|
19 |
"target_loss": null,
|
20 |
"print_eval": true,
|
21 |
"test_delay_epochs": -1,
|
|
|
24 |
"distributed_backend": "nccl",
|
25 |
"distributed_url": "tcp://localhost:54321",
|
26 |
"mixed_precision": true,
|
27 |
+
"precision": "fp16",
|
28 |
+
"epochs": 1300,
|
29 |
+
"batch_size": 48,
|
30 |
"eval_batch_size": 16,
|
31 |
"grad_clip": [
|
32 |
1000,
|
33 |
1000
|
34 |
],
|
35 |
"scheduler_after_epoch": true,
|
36 |
+
"lr": 0.0001,
|
37 |
"optimizer": "AdamW",
|
38 |
"optimizer_params": {
|
39 |
"betas": [
|
|
|
43 |
"eps": 1e-09,
|
44 |
"weight_decay": 0.01
|
45 |
},
|
46 |
+
"lr_scheduler": null,
|
47 |
"lr_scheduler_params": {},
|
48 |
"use_grad_scaler": false,
|
49 |
+
"allow_tf32": false,
|
50 |
"cudnn_enable": true,
|
51 |
"cudnn_deterministic": false,
|
52 |
"cudnn_benchmark": false,
|
|
|
66 |
},
|
67 |
"use_phonemes": false,
|
68 |
"phonemizer": null,
|
69 |
+
"phoneme_language": null,
|
70 |
"compute_input_seq_cache": true,
|
71 |
+
"text_cleaner": null,
|
72 |
"enable_eos_bos_chars": false,
|
73 |
"test_sentences_file": "",
|
74 |
+
"phoneme_cache_path": "/mnt/netapp2/Store_uni/home/usc/ci/ams/experimentos/sabela/normal/phoneme_cache",
|
75 |
"characters": {
|
76 |
"characters_class": "TTS.tts.models.vits.VitsCharacters",
|
77 |
"vocab_dict": null,
|
78 |
+
"pad": "<PAD>",
|
79 |
+
"eos": "<EOS>",
|
80 |
+
"bos": "<BOS>",
|
81 |
+
"blank": "<BLNK>",
|
82 |
+
"characters": "ABCDEFGHIJKLMNOPQRSTUVWXYZ\u00c7\u00c1\u00c9\u00cd\u00d3\u00da\u00dcabcdefghijklmnopqrstuvwxyz\u00f1\u00e1\u00e9\u00ed\u00f3\u00fa\u00fc",
|
83 |
+
"punctuations": "!\u00a1'(),-.:;\u00bf? \"\n",
|
84 |
+
"phonemes": null,
|
85 |
"is_unique": true,
|
86 |
"is_sorted": true
|
87 |
},
|
|
|
93 |
"min_text_len": 1,
|
94 |
"max_text_len": Infinity,
|
95 |
"compute_f0": false,
|
96 |
+
"compute_energy": false,
|
97 |
"compute_linear_spec": true,
|
98 |
"precompute_num_workers": 0,
|
99 |
"start_by_longest": false,
|
100 |
+
"shuffle": false,
|
101 |
+
"drop_last": false,
|
102 |
"datasets": [
|
103 |
{
|
104 |
+
"formatter": "nos_fonemas",
|
105 |
+
"dataset_name": "",
|
106 |
+
"path": "../..",
|
107 |
+
"meta_file_train": "sabela/normal/sabela_train.csv",
|
108 |
"ignored_speakers": null,
|
109 |
"language": "",
|
110 |
+
"phonemizer": "",
|
111 |
"meta_file_val": "",
|
112 |
"meta_file_attn_mask": ""
|
113 |
}
|
114 |
],
|
115 |
"test_sentences": [
|
116 |
+
"k\u00f3mo B\u00e1N eSekut\u00e1r t\u00f3Das \u00e9stas fuNTj\u00f3Ns ke Zes est\u00e1N a peD\u00edr?",
|
117 |
+
"\u00c9 Dif\u00edTil, n\u00f3N?",
|
118 |
+
"a eskas\u00e9Ta De BaT\u00ednas eN eur\u00d3pa poDer\u00eda komplik\u00e1r a aDministraTj\u00f3n da seG\u00fanda D\u00d3se.",
|
119 |
+
"ter\u00e1s \u00e1lGo ke kont\u00e1r...",
|
120 |
+
"k\u00e9 B\u00c9N \u00d3le!",
|
121 |
+
"d\u00d3nas e kaBal\u00e9jros: b\u00f3wZes amos\u00e1lo \u00d3me m\u00e1js f\u00d3rte Do m\u00fando.",
|
122 |
+
"a kur\u00faSa! -berr\u00f3w o B\u00c9Zo.",
|
123 |
+
"as awtoestr\u00e1Das Da S\u00fanta (b\u00edGo Baj\u00f3na e a kor\u00faJa karB\u00e1Zo) estar\u00e1N rremat\u00e1Das no \u00e1no m\u00edl noBeT\u00e9ntos noB\u00e9nta e s\u00e9te.",
|
124 |
+
"\u00e1j!, e SoG\u00e1ndo O tr\u00f3mpo tam\u00c9N aB\u00eda S\u00e9nte ke o koZ\u00eda a \u00faJa, n\u00f3N s\u00e1Bes?, e B\u00c9lo koZ\u00e9r a \u00faJa...",
|
125 |
+
"os ostal\u00e9jros m\u00edraN as T\u00edfras kom preokupaTj\u00f3N."
|
126 |
],
|
127 |
"eval_split_max_size": null,
|
128 |
+
"eval_split_size": 0.05,
|
129 |
"use_speaker_weighted_sampler": false,
|
130 |
"speaker_weighted_sampler_alpha": 1.0,
|
131 |
"use_language_weighted_sampler": false,
|
|
|
133 |
"use_length_weighted_sampler": false,
|
134 |
"length_weighted_sampler_alpha": 1.0,
|
135 |
"model_args": {
|
136 |
+
"num_chars": 83,
|
137 |
"out_channels": 513,
|
138 |
"spec_segment_size": 32,
|
139 |
"hidden_channels": 192,
|
|
|
260 |
"use_d_vector_file": false,
|
261 |
"d_vector_file": null,
|
262 |
"d_vector_dim": 0,
|
263 |
+
"github_branch": "inside_docker"
|
264 |
}
|