Upload folder using huggingface_hub
Browse files- autoregressive.pth +3 -0
- clvp2.pth +3 -0
- config.json +149 -0
- cvvp.pth +3 -0
- diffusion_decoder.pth +3 -0
- mel_norms.pth +3 -0
- rlg_auto.pth +3 -0
- rlg_diffuser.pth +3 -0
- vocoder.pth +3 -0
autoregressive.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:28cd397d84d01a0e01115775f9e9785628b848a889394fc749fa3a232fa5e0d4
|
3 |
+
size 1716990470
|
clvp2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6097e708cf692eb93bd770880660953935e87e8995eb864819bbe51b7d91342c
|
3 |
+
size 975620731
|
config.json
ADDED
@@ -0,0 +1,149 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
{
|
2 |
+
"output_path": "output",
|
3 |
+
"logger_uri": null,
|
4 |
+
"run_name": "run",
|
5 |
+
"project_name": null,
|
6 |
+
"run_description": "\ud83d\udc38Coqui trainer run.",
|
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": 0,
|
21 |
+
"run_eval": true,
|
22 |
+
"run_eval_steps": null,
|
23 |
+
"distributed_backend": "nccl",
|
24 |
+
"distributed_url": "tcp://localhost:54321",
|
25 |
+
"mixed_precision": false,
|
26 |
+
"epochs": 1000,
|
27 |
+
"batch_size": 32,
|
28 |
+
"eval_batch_size": 16,
|
29 |
+
"grad_clip": 0.0,
|
30 |
+
"scheduler_after_epoch": true,
|
31 |
+
"lr": 0.001,
|
32 |
+
"optimizer": "radam",
|
33 |
+
"optimizer_params": null,
|
34 |
+
"lr_scheduler": null,
|
35 |
+
"lr_scheduler_params": {},
|
36 |
+
"use_grad_scaler": false,
|
37 |
+
"cudnn_enable": true,
|
38 |
+
"cudnn_deterministic": false,
|
39 |
+
"cudnn_benchmark": false,
|
40 |
+
"training_seed": 54321,
|
41 |
+
"model": "tortoise",
|
42 |
+
"num_loader_workers": 0,
|
43 |
+
"num_eval_loader_workers": 0,
|
44 |
+
"use_noise_augment": false,
|
45 |
+
"audio": {
|
46 |
+
"sample_rate": 22050,
|
47 |
+
"diffusion_sample_rate": 24000,
|
48 |
+
"output_sample_rate": 24000
|
49 |
+
},
|
50 |
+
"use_phonemes": false,
|
51 |
+
"phonemizer": null,
|
52 |
+
"phoneme_language": null,
|
53 |
+
"compute_input_seq_cache": false,
|
54 |
+
"text_cleaner": null,
|
55 |
+
"enable_eos_bos_chars": false,
|
56 |
+
"test_sentences_file": "",
|
57 |
+
"phoneme_cache_path": null,
|
58 |
+
"characters": null,
|
59 |
+
"add_blank": false,
|
60 |
+
"batch_group_size": 0,
|
61 |
+
"loss_masking": null,
|
62 |
+
"min_audio_len": 1,
|
63 |
+
"max_audio_len": Infinity,
|
64 |
+
"min_text_len": 1,
|
65 |
+
"max_text_len": Infinity,
|
66 |
+
"compute_f0": false,
|
67 |
+
"compute_energy": false,
|
68 |
+
"compute_linear_spec": false,
|
69 |
+
"precompute_num_workers": 0,
|
70 |
+
"start_by_longest": false,
|
71 |
+
"shuffle": false,
|
72 |
+
"drop_last": false,
|
73 |
+
"datasets": [
|
74 |
+
{
|
75 |
+
"formatter": "",
|
76 |
+
"dataset_name": "",
|
77 |
+
"path": "",
|
78 |
+
"meta_file_train": "",
|
79 |
+
"ignored_speakers": null,
|
80 |
+
"language": "",
|
81 |
+
"phonemizer": "",
|
82 |
+
"meta_file_val": "",
|
83 |
+
"meta_file_attn_mask": ""
|
84 |
+
}
|
85 |
+
],
|
86 |
+
"test_sentences": [],
|
87 |
+
"eval_split_max_size": null,
|
88 |
+
"eval_split_size": 0.01,
|
89 |
+
"use_speaker_weighted_sampler": false,
|
90 |
+
"speaker_weighted_sampler_alpha": 1.0,
|
91 |
+
"use_language_weighted_sampler": false,
|
92 |
+
"language_weighted_sampler_alpha": 1.0,
|
93 |
+
"use_length_weighted_sampler": false,
|
94 |
+
"length_weighted_sampler_alpha": 1.0,
|
95 |
+
"model_args": {
|
96 |
+
"autoregressive_batch_size": 1,
|
97 |
+
"enable_redaction": false,
|
98 |
+
"high_vram": false,
|
99 |
+
"kv_cache": true,
|
100 |
+
"ar_checkpoint": null,
|
101 |
+
"clvp_checkpoint": null,
|
102 |
+
"diff_checkpoint": null,
|
103 |
+
"num_chars": 255,
|
104 |
+
"ar_max_mel_tokens": 604,
|
105 |
+
"ar_max_text_tokens": 402,
|
106 |
+
"ar_max_conditioning_inputs": 2,
|
107 |
+
"ar_layers": 30,
|
108 |
+
"ar_model_dim": 1024,
|
109 |
+
"ar_heads": 16,
|
110 |
+
"ar_number_text_tokens": 255,
|
111 |
+
"ar_start_text_token": 255,
|
112 |
+
"ar_checkpointing": false,
|
113 |
+
"ar_train_solo_embeddings": false,
|
114 |
+
"diff_model_channels": 1024,
|
115 |
+
"diff_num_layers": 10,
|
116 |
+
"diff_in_channels": 100,
|
117 |
+
"diff_out_channels": 200,
|
118 |
+
"diff_in_latent_channels": 1024,
|
119 |
+
"diff_in_tokens": 8193,
|
120 |
+
"diff_dropout": 0,
|
121 |
+
"diff_use_fp16": false,
|
122 |
+
"diff_num_heads": 16,
|
123 |
+
"diff_layer_drop": 0,
|
124 |
+
"diff_unconditioned_percentage": 0,
|
125 |
+
"clvp_dim_text": 768,
|
126 |
+
"clvp_dim_speech": 768,
|
127 |
+
"clvp_dim_latent": 768,
|
128 |
+
"clvp_num_text_tokens": 256,
|
129 |
+
"clvp_text_enc_depth": 20,
|
130 |
+
"clvp_text_seq_len": 350,
|
131 |
+
"clvp_text_heads": 12,
|
132 |
+
"clvp_num_speech_tokens": 8192,
|
133 |
+
"clvp_speech_enc_depth": 20,
|
134 |
+
"clvp_speech_heads": 12,
|
135 |
+
"clvp_speech_seq_len": 430,
|
136 |
+
"clvp_use_xformers": true,
|
137 |
+
"duration_const": 102400
|
138 |
+
},
|
139 |
+
"model_dir": null,
|
140 |
+
"temperature": 0.2,
|
141 |
+
"length_penalty": 1.0,
|
142 |
+
"repetition_penalty": 2.0,
|
143 |
+
"top_p": 0.8,
|
144 |
+
"cond_free_k": 2.0,
|
145 |
+
"diffusion_temperature": 1.0,
|
146 |
+
"num_autoregressive_samples": 16,
|
147 |
+
"diffusion_iterations": 30,
|
148 |
+
"sampler": "ddim"
|
149 |
+
}
|
cvvp.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d050e32592ad4a318e03a4f99b09c9c26baf68d78a9d7503ff2bc3883e897100
|
3 |
+
size 151223901
|
diffusion_decoder.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ea776fc354eabb70cfae145777153483fad72e3e0c5ea345505ded2231a90ce1
|
3 |
+
size 1169472627
|
mel_norms.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1f69422a8a8f344c4fca2f0c6b8d41d2151d6615b7321e48e6bb15ae949b119c
|
3 |
+
size 1067
|
rlg_auto.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4473c125482e2a3322a5ea762025a0c6ec657955c3002cf099c0635d79967551
|
3 |
+
size 25193729
|
rlg_diffuser.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6e84b1ce60631c56dc8dec3d27c131993dd99d3060e7919cc351857457dbfdac
|
3 |
+
size 100715777
|
vocoder.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:16e8153e9f8ffb00b116f7f67833df2802fcf81e6bc173acc3b3b4bf9f04189d
|
3 |
+
size 391384715
|