Hugging Face's logo

MegaTTS3
/
diffusion_transformer

Model card Files Community
mrfakename commited on
Commit
284a6e8
·
verified ·
1 Parent(s): 8568b17

Upload folder using huggingface_hub

Browse files
Files changed (2) hide show
  1. config.yaml +148 -0
  2. model_only_last.ckpt +3 -0
config.yaml ADDED
@@ -0,0 +1,148 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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_inp_add_noise: false
33
+ dec_kernel_size: 5
34
+ dec_layers: 4
35
+ dec_post_net_kernel: 3
36
+ decoder_rnn_dim: 0
37
+ decoder_type: conv
38
+ dropout: 0.0
39
+ ds_add_pitch_embed: false
40
+ dur_alpha: 1.0
41
+ dur_context_enc: true
42
+ dur_log: true
43
+ dur_predictor_kernel: 3
44
+ dur_predictor_layers: 2
45
+ dur_use_char: true
46
+ dur_use_spk: true
47
+ enc_dec_norm: ln
48
+ enc_dilations:
49
+ - 1
50
+ - 1
51
+ - 1
52
+ - 1
53
+ enc_ffn_kernel_size: 5
54
+ enc_kernel_size: 5
55
+ enc_layers: 8
56
+ enc_post_net_kernel: 3
57
+ enc_pre_ln: true
58
+ enc_prenet: true
59
+ encoder_K: 8
60
+ encoder_type: rel_fft
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
+ gen_dir_name: ''
71
+ hidden_size: 512
72
+ hop_size: 240
73
+ ignore_begin_end_sil: false
74
+ keep_c0_init: true
75
+ kl_min: 0
76
+ kl_start_steps: 1
77
+ latent_dim: 32
78
+ latent_size: 256
79
+ layers_in_block: 2
80
+ ling_label_dict_size:
81
+ - 20
82
+ - 4
83
+ - 5
84
+ - 2
85
+ - 3
86
+ - 3
87
+ - 3
88
+ - 6
89
+ - 15
90
+ ling_labels:
91
+ - tone
92
+ loud_norm: false
93
+ max_input_tokens: 1550
94
+ mel_vmax: 0.5
95
+ mel_vmin: -6
96
+ min_frames: 50
97
+ mix_melout_timbre: true
98
+ mix_ph_timbre: false
99
+ mixed_precision: bf16
100
+ no_text_enc: false
101
+ num_heads: 2
102
+ out_wav_norm: true
103
+ pad_frames: false
104
+
105
+ precision: bf16
106
+ seed: 1234
107
+ use_bert_input: false
108
+ use_cfg: true
109
+ use_char: true
110
+ use_cur_global: false
111
+ use_cur_global_dec: true
112
+ use_dit_1b: false
113
+ use_dur_embed: true
114
+ use_dur_mask_embed: true
115
+ use_ema: false
116
+ use_expand_ph: true
117
+ use_finegrained_spk: false
118
+ use_global_lat: false
119
+ use_gt_dur: false
120
+ use_gt_f0: false
121
+ use_mix_spk_embed: false
122
+ use_new_vae: false
123
+ use_ph_level_f0: false
124
+ use_ph_pos_embed: true
125
+ use_pitch_embed: false
126
+ use_pitch_embed_dec: false
127
+ use_pitch_pred: true
128
+ use_pos_embed: true
129
+ use_qk_norm: true
130
+ use_random_spk_embed: false
131
+ use_seq_cfg: true
132
+ use_spk_embed: false
133
+ use_spk_enc: true
134
+ use_spk_id: false
135
+ use_uv: true
136
+ use_vae: true
137
+ use_vpcfm: true
138
+ use_vqvae: true
139
+ use_word_encoder: true
140
+ use_word_input: false
141
+ vae_dur_grad: 0.1
142
+ vae_enc_hidden_size: 384
143
+ vae_stride: 4
144
+ vae_word_conder_layers: 0
145
+ vq_stride: 8
146
+ vqvae_start_steps: 0
147
+ win_size: 1200
148
+ word_dict_size: 10000
model_only_last.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12233b95be177504551034390cf71aa748f0c66cbe2fd0ce433b9f9686122da9
3
+ size 1836341777