Hugging Face's logo

MegaTTS3
/
aligner_lm

Model card Files Community
mrfakename commited on
Commit
4c08523
·
verified ·
1 Parent(s): b6c8464

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. config.yaml +151 -0
  2. model_only_last.ckpt +3 -0
config.yaml ADDED
@@ -0,0 +1,151 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ acous_params:
2
+ - - 480
3
+ - 1200
4
+ - 80
5
+ - - 240
6
+ - 1200
7
+ - 160
8
+ amp: true
9
+ audio_num_mel_bins: 160
10
+ audio_sample_rate: 24000
11
+ base_config:
12
+ - ./base_config.yaml
13
+ c_spk_enc: 512
14
+ char_dict_size: 15000
15
+ conv_use_pos: false
16
+ dec0_dilations:
17
+ - 1
18
+ - 2
19
+ - 4
20
+ - 1
21
+ - 2
22
+ - 4
23
+ - 1
24
+ dec0_kernel_size: 3
25
+ dec_dilations:
26
+ - 1
27
+ - 2
28
+ - 1
29
+ - 2
30
+ - 1
31
+ dec_ffn_kernel_size: 9
32
+ dec_kernel_size: 5
33
+ dec_layers: 4
34
+ dec_post_net_kernel: 3
35
+ decoder_rnn_dim: 0
36
+ decoder_type: conv
37
+ dropout: 0.0
38
+ dur_alpha: 1.0
39
+ dur_context_enc: true
40
+ dur_log: true
41
+ dur_predictor_kernel: 3
42
+ dur_predictor_layers: 2
43
+ dur_use_char: true
44
+ dur_use_spk: true
45
+ enc_dec_norm: ln
46
+ enc_dilations:
47
+ - 1
48
+ - 1
49
+ - 1
50
+ - 1
51
+ enc_ffn_kernel_size: 5
52
+ enc_kernel_size: 5
53
+ enc_layers: 8
54
+ enc_post_net_kernel: 3
55
+ enc_pre_ln: true
56
+ enc_prenet: true
57
+ encoder_K: 8
58
+ encoder_type: rel_fft
59
+ endless_ds: true
60
+ eval_max_batches: 0
61
+ f0_max: 600
62
+ f0_min: 60
63
+ ffn_act: gelu
64
+ ffn_hidden_size: 1024
65
+ fft_size: 1200
66
+ fg_spk_enc_hidden: 256
67
+ fmax: 12000
68
+ fmin: 0
69
+ frames_multiple: 8
70
+ hidden_size: 512
71
+ hop_size: 240
72
+ keep_c0_init: true
73
+ lat_for_dur: false
74
+ latent_dim: 16
75
+ latent_size: 256
76
+ layers_in_block: 2
77
+ ling_label_dict_size:
78
+ - 20
79
+ - 4
80
+ - 5
81
+ - 2
82
+ - 3
83
+ - 3
84
+ - 3
85
+ - 6
86
+ - 15
87
+ ling_labels:
88
+ - tone
89
+ loud_norm: false
90
+ mel_vmax: 0.5
91
+ mel_vmin: -6
92
+ min_frames: 50
93
+ mix_melout_timbre: true
94
+ mix_ph_timbre: false
95
+ mixed_precision: bf16
96
+ model_type: 1
97
+ multistage: false
98
+ no_text_enc: false
99
+ num_ckpt_keep: 5
100
+ num_heads: 2
101
+ num_spk: 50000
102
+ out_wav_norm: true
103
+ pitch_extractor: reaper
104
+ pitch_key: pitch
105
+ pitch_type: frame
106
+ precision: bf16
107
+ ref_mel_bins: 160
108
+ seed: 1234
109
+ split_ref: true
110
+ use_bert_input: false
111
+ use_cfg: true
112
+ use_char: true
113
+ use_cur_global: false
114
+ use_cur_global_dec: true
115
+ use_dur_embed: true
116
+ use_dur_mask_embed: true
117
+ use_ema: false
118
+ use_expand_ph: true
119
+ use_finegrained_spk: false
120
+ use_gt_dur: false
121
+ use_gt_f0: false
122
+ use_mix_spk_embed: false
123
+ use_new_vae: false
124
+ use_ph_level_f0: false
125
+ use_ph_pos_embed: true
126
+ use_pitch_embed: false
127
+ use_pitch_embed_dec: false
128
+ use_pitch_pred: true
129
+ use_pos_embed: true
130
+ use_qk_norm: true
131
+ use_random_spk_embed: false
132
+ use_seq_cfg: true
133
+ use_spk_embed: false
134
+ use_spk_enc: true
135
+ use_spk_id: false
136
+ use_uv: true
137
+ use_vae: true
138
+ use_vpcfm: true
139
+ use_vqvae: true
140
+ use_word_encoder: true
141
+ use_word_input: false
142
+ vae_dur_grad: 0.1
143
+ vae_enc_hidden_size: 384
144
+ vae_stride: 4
145
+ vae_word_conder_layers: 0
146
+ vq_stride: 8
147
+ vqvae_start_steps: 0
148
+ win_size: 1200
149
+ word_dict_size: 10000
150
+ z_channels: 64
151
+ z_clamp: 2.0
model_only_last.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a00f18ec36f8c1328ddab7a405c8e388790a1c14fdbdd07c546fcacaf5d19296
3
+ size 218434266