Infinix-io
/

Infinix-io gabrielclark3330 commited on
Commit
ab402f8
·
0 Parent(s):

Duplicate from Zyphra/ZONOS3-Mini

Browse files

Co-authored-by: Gabriel Clark <gabrielclark3330@users.noreply.huggingface.co>

This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. .gitattributes +164 -0
  2. README.md +204 -0
  3. assets/arch_zac1_dark.png +3 -0
  4. assets/eval_pareto.png +3 -0
  5. assets/zonos3_mini_arloop_animated.gif +3 -0
  6. assets/zonos3_mini_hero.webp +3 -0
  7. autoencoder/generator.pt +3 -0
  8. autoencoder/heads/full_mr/config_full_unfreezeall_mr.json +455 -0
  9. autoencoder/heads/full_mr/step_0050000.pt +3 -0
  10. autoencoder/heads/keep2_plain/config_full_unfreezeall_mr_drop30_4x10.json +473 -0
  11. autoencoder/heads/keep2_plain/step_0050000.pt +3 -0
  12. autoencoder/teacher/config.json +81 -0
  13. autoencoder/teacher/model.safetensors +3 -0
  14. gguf/backbone-ptq-edit.gguf +3 -0
  15. gguf/backbone-ptq.gguf +3 -0
  16. gguf/edit/accents.gguf +3 -0
  17. gguf/spk-encoder.gguf +3 -0
  18. gguf/ssl-teacher.gguf +3 -0
  19. gguf/steer.gguf +3 -0
  20. gguf/voices/English/Cillian_en-Irish_M_1.codes.npy +3 -0
  21. gguf/voices/English/Cillian_en-Irish_M_1.embedding.npy +3 -0
  22. gguf/voices/English/Cillian_en-Irish_M_1.wav +3 -0
  23. gguf/voices/English/Imogen_en-British_F_1.codes.npy +3 -0
  24. gguf/voices/English/Imogen_en-British_F_1.embedding.npy +3 -0
  25. gguf/voices/English/Imogen_en-British_F_1.wav +3 -0
  26. gguf/voices/English/Jett_en-energetic_M_1.codes.npy +3 -0
  27. gguf/voices/English/Jett_en-energetic_M_1.embedding.npy +3 -0
  28. gguf/voices/English/Jett_en-energetic_M_1.wav +3 -0
  29. gguf/voices/English/Morgana_en-witch_F_1.codes.npy +3 -0
  30. gguf/voices/English/Morgana_en-witch_F_1.embedding.npy +3 -0
  31. gguf/voices/English/Morgana_en-witch_F_1.mp3 +3 -0
  32. gguf/voices/English/Rupert_en-British_M_1.codes.npy +3 -0
  33. gguf/voices/English/Rupert_en-British_M_1.embedding.npy +3 -0
  34. gguf/voices/English/Rupert_en-British_M_1.wav +3 -0
  35. gguf/voices/English/Sakura_en-anime_F_1.codes.npy +3 -0
  36. gguf/voices/English/Sakura_en-anime_F_1.embedding.npy +3 -0
  37. gguf/voices/English/Sakura_en-anime_F_1.wav +3 -0
  38. gguf/voices/English/Savannah_en-American_F_1.codes.npy +3 -0
  39. gguf/voices/English/Savannah_en-American_F_1.embedding.npy +3 -0
  40. gguf/voices/English/Savannah_en-American_F_1.mp3 +3 -0
  41. gguf/voices/English/Thaddeus_en-announcervoice_M_1.codes.npy +3 -0
  42. gguf/voices/English/Thaddeus_en-announcervoice_M_1.embedding.npy +3 -0
  43. gguf/voices/English/Thaddeus_en-announcervoice_M_1.mp3 +3 -0
  44. gguf/voices/English/Willow_en-asrm_F_1.codes.npy +3 -0
  45. gguf/voices/English/Willow_en-asrm_F_1.embedding.npy +3 -0
  46. gguf/voices/English/Willow_en-asrm_F_1.wav +3 -0
  47. gguf/voices/English/Wyatt_en-American_M_1.codes.npy +3 -0
  48. gguf/voices/English/Wyatt_en-American_M_1.embedding.npy +3 -0
  49. gguf/voices/English/Wyatt_en-American_M_1.ogg +3 -0
  50. gguf/voices/French/Gaspard_fr_M_1.codes.npy +3 -0
.gitattributes ADDED
@@ -0,0 +1,164 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ *.7z filter=lfs diff=lfs merge=lfs -text
2
+ *.arrow filter=lfs diff=lfs merge=lfs -text
3
+ *.bin filter=lfs diff=lfs merge=lfs -text
4
+ *.bz2 filter=lfs diff=lfs merge=lfs -text
5
+ *.ckpt filter=lfs diff=lfs merge=lfs -text
6
+ *.ftz filter=lfs diff=lfs merge=lfs -text
7
+ *.gz filter=lfs diff=lfs merge=lfs -text
8
+ *.h5 filter=lfs diff=lfs merge=lfs -text
9
+ *.joblib filter=lfs diff=lfs merge=lfs -text
10
+ *.lfs.* filter=lfs diff=lfs merge=lfs -text
11
+ *.mlmodel filter=lfs diff=lfs merge=lfs -text
12
+ *.model filter=lfs diff=lfs merge=lfs -text
13
+ *.msgpack filter=lfs diff=lfs merge=lfs -text
14
+ *.npy filter=lfs diff=lfs merge=lfs -text
15
+ *.npz filter=lfs diff=lfs merge=lfs -text
16
+ *.onnx filter=lfs diff=lfs merge=lfs -text
17
+ *.ot filter=lfs diff=lfs merge=lfs -text
18
+ *.parquet filter=lfs diff=lfs merge=lfs -text
19
+ *.pb filter=lfs diff=lfs merge=lfs -text
20
+ *.pickle filter=lfs diff=lfs merge=lfs -text
21
+ *.pkl filter=lfs diff=lfs merge=lfs -text
22
+ *.pt filter=lfs diff=lfs merge=lfs -text
23
+ *.pth filter=lfs diff=lfs merge=lfs -text
24
+ *.rar filter=lfs diff=lfs merge=lfs -text
25
+ *.safetensors filter=lfs diff=lfs merge=lfs -text
26
+ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
27
+ *.tar.* filter=lfs diff=lfs merge=lfs -text
28
+ *.tar filter=lfs diff=lfs merge=lfs -text
29
+ *.tflite filter=lfs diff=lfs merge=lfs -text
30
+ *.tgz filter=lfs diff=lfs merge=lfs -text
31
+ *.wasm filter=lfs diff=lfs merge=lfs -text
32
+ *.xz filter=lfs diff=lfs merge=lfs -text
33
+ *.zip filter=lfs diff=lfs merge=lfs -text
34
+ *.zst filter=lfs diff=lfs merge=lfs -text
35
+ *tfevents* filter=lfs diff=lfs merge=lfs -text
36
+ assets/eval_itw.png filter=lfs diff=lfs merge=lfs -text
37
+ assets/eval_clean.png filter=lfs diff=lfs merge=lfs -text
38
+ assets/eval_ranking.png filter=lfs diff=lfs merge=lfs -text
39
+ assets/arch_zonos3_mini.png filter=lfs diff=lfs merge=lfs -text
40
+ assets/arch_zac1.png filter=lfs diff=lfs merge=lfs -text
41
+ assets/zonos3_mini_arloop_animated.gif filter=lfs diff=lfs merge=lfs -text
42
+ assets/arch_zac1_dark.png filter=lfs diff=lfs merge=lfs -text
43
+ assets/eval_pareto.png filter=lfs diff=lfs merge=lfs -text
44
+ assets/zonos3_mini_hero.webp filter=lfs diff=lfs merge=lfs -text
45
+ gguf/edit/accents.gguf filter=lfs diff=lfs merge=lfs -text
46
+ gguf/spk-encoder.gguf filter=lfs diff=lfs merge=lfs -text
47
+ gguf/ssl-teacher.gguf filter=lfs diff=lfs merge=lfs -text
48
+ gguf/steer.gguf filter=lfs diff=lfs merge=lfs -text
49
+ gguf/voices/Ananya_hi_F_1.wav filter=lfs diff=lfs merge=lfs -text
50
+ gguf/voices/Arjun_hi_M_1.wav filter=lfs diff=lfs merge=lfs -text
51
+ gguf/voices/Chaitanya_te_M_1.wav filter=lfs diff=lfs merge=lfs -text
52
+ gguf/voices/Cillian_en-Irish_M_1.wav filter=lfs diff=lfs merge=lfs -text
53
+ gguf/voices/Gaspard_fr_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
54
+ gguf/voices/Ginevra_it_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
55
+ gguf/voices/Haoran_zh-energetic_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
56
+ gguf/voices/Imogen_en-British_F_1.wav filter=lfs diff=lfs merge=lfs -text
57
+ gguf/voices/Jett_en-energetic_M_1.wav filter=lfs diff=lfs merge=lfs -text
58
+ gguf/voices/Leonor_pt_F_1.wav filter=lfs diff=lfs merge=lfs -text
59
+ gguf/voices/Liesel_de_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
60
+ gguf/voices/Lorenzo_it_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
61
+ gguf/voices/Meilin_zh-energetic_F_1.wav filter=lfs diff=lfs merge=lfs -text
62
+ gguf/voices/Minjun_ko_M_1.wav filter=lfs diff=lfs merge=lfs -text
63
+ gguf/voices/Morgana_en-witch_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
64
+ gguf/voices/Rodrigo_es_M_1.wav filter=lfs diff=lfs merge=lfs -text
65
+ gguf/voices/Rupert_en-British_M_1.wav filter=lfs diff=lfs merge=lfs -text
66
+ gguf/voices/Sakura_en-anime_F_1.wav filter=lfs diff=lfs merge=lfs -text
67
+ gguf/voices/Savannah_en-American_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
68
+ gguf/voices/Seoyeon_ko_F_1.wav filter=lfs diff=lfs merge=lfs -text
69
+ gguf/voices/Sravya_te_F_1.wav filter=lfs diff=lfs merge=lfs -text
70
+ gguf/voices/Thaddeus_en-announcervoice_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
71
+ gguf/voices/Thiago_pt_M_1.wav filter=lfs diff=lfs merge=lfs -text
72
+ gguf/voices/Willow_en-asrm_F_1.wav filter=lfs diff=lfs merge=lfs -text
73
+ gguf/voices/Wolfgang_de_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
74
+ gguf/voices/Wyatt_en-American_M_1.ogg filter=lfs diff=lfs merge=lfs -text
75
+ gguf/voices/Ximena_es_F_1.wav filter=lfs diff=lfs merge=lfs -text
76
+ gguf/zac-encoder.gguf filter=lfs diff=lfs merge=lfs -text
77
+ gguf/zac-keep2.gguf filter=lfs diff=lfs merge=lfs -text
78
+ gguf/zac.gguf filter=lfs diff=lfs merge=lfs -text
79
+ gguf/zonos3-ptq-edit.gguf filter=lfs diff=lfs merge=lfs -text
80
+ gguf/zonos3-ptq.gguf filter=lfs diff=lfs merge=lfs -text
81
+ web/companions/steer.gguf filter=lfs diff=lfs merge=lfs -text
82
+ web/voices/Ananya_hi_F_1.wav filter=lfs diff=lfs merge=lfs -text
83
+ web/voices/Arjun_hi_M_1.wav filter=lfs diff=lfs merge=lfs -text
84
+ web/voices/Chaitanya_te_M_1.wav filter=lfs diff=lfs merge=lfs -text
85
+ web/voices/Cillian_en-Irish_M_1.wav filter=lfs diff=lfs merge=lfs -text
86
+ web/voices/Gaspard_fr_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
87
+ web/voices/Ginevra_it_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
88
+ web/voices/Haoran_zh-energetic_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
89
+ web/voices/Imogen_en-British_F_1.wav filter=lfs diff=lfs merge=lfs -text
90
+ web/voices/Jett_en-energetic_M_1.wav filter=lfs diff=lfs merge=lfs -text
91
+ web/voices/Leonor_pt_F_1.wav filter=lfs diff=lfs merge=lfs -text
92
+ web/voices/Liesel_de_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
93
+ web/voices/Lorenzo_it_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
94
+ web/voices/Meilin_zh-energetic_F_1.wav filter=lfs diff=lfs merge=lfs -text
95
+ web/voices/Minjun_ko_M_1.wav filter=lfs diff=lfs merge=lfs -text
96
+ web/voices/Morgana_en-witch_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
97
+ web/voices/Rodrigo_es_M_1.wav filter=lfs diff=lfs merge=lfs -text
98
+ web/voices/Rupert_en-British_M_1.wav filter=lfs diff=lfs merge=lfs -text
99
+ web/voices/Sakura_en-anime_F_1.wav filter=lfs diff=lfs merge=lfs -text
100
+ web/voices/Savannah_en-American_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
101
+ web/voices/Seoyeon_ko_F_1.wav filter=lfs diff=lfs merge=lfs -text
102
+ web/voices/Sravya_te_F_1.wav filter=lfs diff=lfs merge=lfs -text
103
+ web/voices/Thaddeus_en-announcervoice_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
104
+ web/voices/Thiago_pt_M_1.wav filter=lfs diff=lfs merge=lfs -text
105
+ web/voices/Willow_en-asrm_F_1.wav filter=lfs diff=lfs merge=lfs -text
106
+ web/voices/Wolfgang_de_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
107
+ web/voices/Wyatt_en-American_M_1.ogg filter=lfs diff=lfs merge=lfs -text
108
+ web/voices/Ximena_es_F_1.wav filter=lfs diff=lfs merge=lfs -text
109
+ gguf/voices/English/Cillian_en-Irish_M_1.wav filter=lfs diff=lfs merge=lfs -text
110
+ gguf/voices/English/Imogen_en-British_F_1.wav filter=lfs diff=lfs merge=lfs -text
111
+ gguf/voices/English/Jett_en-energetic_M_1.wav filter=lfs diff=lfs merge=lfs -text
112
+ gguf/voices/English/Morgana_en-witch_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
113
+ gguf/voices/English/Rupert_en-British_M_1.wav filter=lfs diff=lfs merge=lfs -text
114
+ gguf/voices/English/Sakura_en-anime_F_1.wav filter=lfs diff=lfs merge=lfs -text
115
+ gguf/voices/English/Savannah_en-American_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
116
+ gguf/voices/English/Thaddeus_en-announcervoice_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
117
+ gguf/voices/English/Willow_en-asrm_F_1.wav filter=lfs diff=lfs merge=lfs -text
118
+ gguf/voices/English/Wyatt_en-American_M_1.ogg filter=lfs diff=lfs merge=lfs -text
119
+ gguf/voices/French/Gaspard_fr_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
120
+ gguf/voices/German/Liesel_de_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
121
+ gguf/voices/German/Wolfgang_de_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
122
+ gguf/voices/Hindi/Ananya_hi_F_1.wav filter=lfs diff=lfs merge=lfs -text
123
+ gguf/voices/Hindi/Arjun_hi_M_1.wav filter=lfs diff=lfs merge=lfs -text
124
+ gguf/voices/Italian/Ginevra_it_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
125
+ gguf/voices/Italian/Lorenzo_it_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
126
+ gguf/voices/Korean/Minjun_ko_M_1.wav filter=lfs diff=lfs merge=lfs -text
127
+ gguf/voices/Korean/Seoyeon_ko_F_1.wav filter=lfs diff=lfs merge=lfs -text
128
+ gguf/voices/Mandarin[[:space:]]Chinese/Haoran_zh-energetic_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
129
+ gguf/voices/Mandarin[[:space:]]Chinese/Meilin_zh-energetic_F_1.wav filter=lfs diff=lfs merge=lfs -text
130
+ gguf/voices/Portuguese/Leonor_pt_F_1.wav filter=lfs diff=lfs merge=lfs -text
131
+ gguf/voices/Portuguese/Thiago_pt_M_1.wav filter=lfs diff=lfs merge=lfs -text
132
+ gguf/voices/Spanish/Rodrigo_es_M_1.wav filter=lfs diff=lfs merge=lfs -text
133
+ gguf/voices/Spanish/Ximena_es_F_1.wav filter=lfs diff=lfs merge=lfs -text
134
+ gguf/voices/Telugu/Chaitanya_te_M_1.wav filter=lfs diff=lfs merge=lfs -text
135
+ gguf/voices/Telugu/Sravya_te_F_1.wav filter=lfs diff=lfs merge=lfs -text
136
+ web/voices/English/Cillian_en-Irish_M_1.wav filter=lfs diff=lfs merge=lfs -text
137
+ web/voices/English/Imogen_en-British_F_1.wav filter=lfs diff=lfs merge=lfs -text
138
+ web/voices/English/Jett_en-energetic_M_1.wav filter=lfs diff=lfs merge=lfs -text
139
+ web/voices/English/Morgana_en-witch_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
140
+ web/voices/English/Rupert_en-British_M_1.wav filter=lfs diff=lfs merge=lfs -text
141
+ web/voices/English/Sakura_en-anime_F_1.wav filter=lfs diff=lfs merge=lfs -text
142
+ web/voices/English/Savannah_en-American_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
143
+ web/voices/English/Thaddeus_en-announcervoice_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
144
+ web/voices/English/Willow_en-asrm_F_1.wav filter=lfs diff=lfs merge=lfs -text
145
+ web/voices/English/Wyatt_en-American_M_1.ogg filter=lfs diff=lfs merge=lfs -text
146
+ web/voices/French/Gaspard_fr_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
147
+ web/voices/German/Liesel_de_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
148
+ web/voices/German/Wolfgang_de_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
149
+ web/voices/Hindi/Ananya_hi_F_1.wav filter=lfs diff=lfs merge=lfs -text
150
+ web/voices/Hindi/Arjun_hi_M_1.wav filter=lfs diff=lfs merge=lfs -text
151
+ web/voices/Italian/Ginevra_it_F_1.mp3 filter=lfs diff=lfs merge=lfs -text
152
+ web/voices/Italian/Lorenzo_it_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
153
+ web/voices/Korean/Minjun_ko_M_1.wav filter=lfs diff=lfs merge=lfs -text
154
+ web/voices/Korean/Seoyeon_ko_F_1.wav filter=lfs diff=lfs merge=lfs -text
155
+ web/voices/Mandarin[[:space:]]Chinese/Haoran_zh-energetic_M_1.mp3 filter=lfs diff=lfs merge=lfs -text
156
+ web/voices/Mandarin[[:space:]]Chinese/Meilin_zh-energetic_F_1.wav filter=lfs diff=lfs merge=lfs -text
157
+ web/voices/Portuguese/Leonor_pt_F_1.wav filter=lfs diff=lfs merge=lfs -text
158
+ web/voices/Portuguese/Thiago_pt_M_1.wav filter=lfs diff=lfs merge=lfs -text
159
+ web/voices/Spanish/Rodrigo_es_M_1.wav filter=lfs diff=lfs merge=lfs -text
160
+ web/voices/Spanish/Ximena_es_F_1.wav filter=lfs diff=lfs merge=lfs -text
161
+ web/voices/Telugu/Chaitanya_te_M_1.wav filter=lfs diff=lfs merge=lfs -text
162
+ web/voices/Telugu/Sravya_te_F_1.wav filter=lfs diff=lfs merge=lfs -text
163
+ gguf/backbone-ptq.gguf filter=lfs diff=lfs merge=lfs -text
164
+ gguf/backbone-ptq-edit.gguf filter=lfs diff=lfs merge=lfs -text
README.md ADDED
@@ -0,0 +1,204 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ pipeline_tag: text-to-speech
4
+ library_name: zonos3
5
+ language:
6
+ - en
7
+ - fr
8
+ - es
9
+ - de
10
+ - it
11
+ - pt
12
+ - nl
13
+ - pl
14
+ - ru
15
+ - sv
16
+ - tr
17
+ - uk
18
+ - cs
19
+ - da
20
+ - fi
21
+ - el
22
+ - hu
23
+ - "no"
24
+ - ro
25
+ - sk
26
+ - bg
27
+ - zh
28
+ - ja
29
+ - ko
30
+ - ar
31
+ - he
32
+ - vi
33
+ - id
34
+ - th
35
+ - tl
36
+ - ms
37
+ - ka
38
+ - hi
39
+ - bn
40
+ - ta
41
+ - te
42
+ - mr
43
+ - gu
44
+ - kn
45
+ - ml
46
+ - pa
47
+ - or
48
+ - ur
49
+ tags:
50
+ - text-to-speech
51
+ - tts
52
+ - speech-synthesis
53
+ - voice-cloning
54
+ - multilingual
55
+ - gguf
56
+ - zonos
57
+ - zonos3
58
+ - zyphra
59
+ ---
60
+
61
+ # ZONOS3-Mini
62
+
63
+ <p align="center">
64
+ <img src="./assets/zonos3_mini_hero.webp" alt="ZONOS3-Mini title card" width="750" />
65
+ </p>
66
+
67
+ <div align="center">
68
+ <a href="https://arxiv.org/abs/2611.16525" target="_blank">
69
+ <img src="https://img.shields.io/badge/arXiv-2611.16525-B31B1B?style=for-the-badge&logo=arxiv&logoColor=white" alt="arXiv">
70
+ </a>
71
+ <a href="https://github.com/Zyphra/ZONOS3-Mini" target="_blank">
72
+ <img src="https://img.shields.io/badge/GitHub-ZONOS3--Mini-181717?style=for-the-badge&logo=github&logoColor=white" alt="GitHub">
73
+ </a>
74
+ <a href="https://discord.gg/gTW9JwST8q" target="_blank">
75
+ <img src="https://img.shields.io/badge/Join%20Our%20Discord-7289DA?style=for-the-badge&logo=discord&logoColor=white" alt="Discord">
76
+ </a>
77
+ </div>
78
+
79
+ ---
80
+
81
+ ZONOS3-Mini is Zyphra’s compact, open-source text-to-speech model, delivering natural reliable speech and high-fidelity voice
82
+ cloning with just 100 million active parameters. It runs in real time on phones and laptops, including directly in the browser.
83
+
84
+ Clone a voice from an audio sample, or design a character by shaping age, accent and vocal texture. Steer emotion, pace and
85
+ delivery down to individual words for games, narration and voice assistants.
86
+
87
+ Run ZONOS3-Mini locally with our browser and native runtimes, or access it through [Zyphra Cloud](https://cloud.zyphra.com/models/zyphra/ZONOS3-Mini). Model weights are released
88
+ under Apache 2.0.
89
+
90
+ **Model details, samples and evaluations are in the [blog post](https://www.zyphra.com/our-work/zonos3-mini).**
91
+ **A hosted version is available at [Zyphra Cloud](https://cloud.zyphra.com/models/zyphra/ZONOS3-Mini).**
92
+
93
+ ## Languages
94
+
95
+ ZONOS3-Mini supports 43 languages.
96
+
97
+ | Region | Languages |
98
+ |---|---|
99
+ | Western and Southern Europe | English, French, Spanish, German, Italian, Portuguese, Dutch, Greek |
100
+ | Northern Europe | Swedish, Danish, Norwegian, Finnish |
101
+ | Central and Eastern Europe | Polish, Czech, Slovak, Hungarian, Romanian, Bulgarian, Russian, Ukrainian |
102
+ | Middle East and Caucasus | Turkish, Arabic, Hebrew, Georgian |
103
+ | East Asia | Mandarin Chinese, Japanese, Korean |
104
+ | Southeast Asia | Vietnamese, Indonesian, Malay, Thai, Tagalog |
105
+ | South Asia | Hindi, Bengali, Tamil, Telugu, Marathi, Gujarati, Kannada, Malayalam, Punjabi, Odia, Urdu |
106
+
107
+ ## Quick start
108
+
109
+ ### In your browser
110
+
111
+ ZONOS3-Mini runs entirely on your own GPU through WebGPU: no install and no server. The first visit downloads the model, and
112
+ the browser keeps it cached for later visits. You need a browser with WebGPU support.
113
+
114
+ - [Zonos Voice Studio](https://zyphra.com/zonos3-mini/localplayground#cast): the full studio.
115
+ - [Lite](https://zyphra.com/zonos3-mini/localplayground/lite): quick voice cloning. Pick, upload or record a voice and
116
+ speak with it.
117
+
118
+ ### With SGLang
119
+
120
+ [SGLang-Omni](https://github.com/sgl-project/sglang-omni) serves ZONOS3-Mini on NVIDIA GPUs behind the OpenAI-compatible
121
+ `/v1/audio/speech` endpoint, with streaming and voice cloning from a reference clip (no transcript needed). Install SGLang-Omni
122
+ by following its [installation guide](https://github.com/sgl-project/sglang-omni/blob/main/docs/get_started/installation.md),
123
+ then start a server:
124
+
125
+ ```bash
126
+ sgl-omni serve --model-path Zyphra/ZONOS3-Mini --port 8000
127
+ ```
128
+
129
+ Once `curl http://localhost:8000/health` reports `"healthy"`, generate speech with the default voice from another terminal:
130
+
131
+ ```bash
132
+ curl -X POST http://localhost:8000/v1/audio/speech \
133
+ -H "Content-Type: application/json" \
134
+ -d '{"input": "Hello from ZONOS3-Mini!"}' \
135
+ --output output.wav
136
+ ```
137
+
138
+ Add `"references": [{"audio_path": "<reference clip>"}]` to clone a voice, or `"stream": true, "response_format": "pcm"` to
139
+ stream 44.1 kHz audio as it is generated. The [ZONOS3 cookbook](https://github.com/sgl-project/sglang-omni/blob/main/docs/cookbook/zonos3.md)
140
+ covers voice cloning, streaming and the sampling options.
141
+
142
+ ### Native runtimes
143
+
144
+ The [ZONOS3-Mini GitHub repository](https://github.com/Zyphra/ZONOS3-Mini) holds Zyphra's own runtimes, which share one HTTP
145
+ API and serve the Zonos Voice Studio: a CUDA server for NVIDIA GPUs, an MLX server for Apple Silicon, a C++/ggml server and CLI
146
+ (CPU, CUDA, Metal or Vulkan), and the in-browser WebGPU build.
147
+ They load the GGUF files in [`gguf/`](https://huggingface.co/Zyphra/ZONOS3-Mini/tree/main/gguf) and the WebGPU shards in
148
+ [`web/`](https://huggingface.co/Zyphra/ZONOS3-Mini/tree/main/web); the 27 default voices, grouped by language, are in
149
+ [`gguf/voices/`](https://huggingface.co/Zyphra/ZONOS3-Mini/blob/main/gguf/voices/README.md).
150
+
151
+ ## Architecture (rough outline)
152
+
153
+ Text is NeMo-normalized (English, Mandarin, Japanese, Korean, German, Spanish, French, Italian, Portuguese; other languages
154
+ pass through as written) and encoded with an 8,192-entry SuperBPE tokenizer. The language model autoregressively generates 15 Hz
155
+ dual-codec frames (10 codebooks of 4,096 entries each: one semantic codebook distilled from the Wav2Vec2-BERT teacher plus nine
156
+ acoustic codebooks), conditioned on the text, a speaker prompt and three bucketed acoustic controls. A causal multi-resolution
157
+ iSTFT head renders the frames to 44.1 kHz audio as they are produced.
158
+
159
+ ![ZONOS3-Mini inference pipeline](assets/zonos3_mini_arloop_animated.gif)
160
+
161
+ *Inference pipeline. Speaker conditioning (raw reference codes or a speaker-embedding row), control rows and SuperBPE text enter
162
+ the hybrid GDN/attention MoE backbone, which predicts all ten codebooks of the next frame under the delay pattern (four drawn);
163
+ as soon as a frame's last codebook is generated, the codec's causal iSTFT head renders that frame to audio.*
164
+
165
+ | | ZONOS3-Mini |
166
+ |---|---|
167
+ | Backbone | 11 layers, model dim 1,024; 800M parameters total, about 100M active per token (latent MoE: 64 experts, top-2 routing, from layer 1) |
168
+ | Sequence mixer | pattern `gdn3_global1`: 9 Gated DeltaNet (GDN) layers + 2 global attention layers (16 heads x 32 dims, 4 KV heads), attention residuals (`full_attnres`); context 32,768 positions |
169
+ | Text | SuperBPE, 8,192 entries (+4 specials), NeMo text normalization at inference |
170
+ | Audio tokens | 15 Hz, 44.1 kHz dual-encoding codec, 10 x 4,096 codebooks (1 semantic + 9 acoustic); hop 2,940 samples; up to 10 minutes of audio per generation |
171
+ | Speaker prompt | raw reference clip (encoded by the codec into a token prefix, up to 60 s, with speaker CFG) **or** a single 192-d MECT-B2 speaker embedding (`ownmect_s2`) |
172
+ | Controls | loudness (LUFS, 10 buckets), bandwidth f_max (Hz, 10 buckets), **speaking rate (UTF-8 bytes per second, 30 buckets)**; each row has an "unknown" index; bucket edges in `params.json` (`condition_edges`) |
173
+ | Decoder | causal multi-resolution iSTFT head: `full_mr` (default) or `keep2_plain` (pruned, about 3x faster decode) |
174
+ | Precision | bf16 weights (`model.pth`); decoder heads run in fp32 |
175
+
176
+ ![ZAC 1 audio codec](assets/arch_zac1_dark.png)
177
+
178
+ *ZAC 1 (Zyphra Audio Codec 1).*
179
+
180
+ ## Evaluations
181
+
182
+ Full results and the ZTTS2 capability index are in the [blog post](https://www.zyphra.com/our-work/zonos3-mini#evaluation).
183
+
184
+ Quality against speed on one H100 80GB. Quality combines macro speaker similarity and macro WER over the 55 language subsets of
185
+ [ZTTS2-Eval](https://zyphra.github.io/ZTTS2-Eval/) (benchmark version 2: FLEURS-R read speech in 35 languages and VoxBlink2 in-the-wild recordings in 20 languages) as
186
+ ½ z(similarity) − ½ z(WER) across the systems shown; ZONOS3-Mini is measured with the speaker-embedding prompt. Left: median
187
+ time to first audio with 8 concurrent requests (80 ms for ZONOS3-Mini). Right: peak throughput at each model's best concurrency
188
+ (457 seconds of audio per second for ZONOS3-Mini).
189
+
190
+ ![ZONOS3-Mini quality versus time to first audio and throughput](assets/eval_pareto.png)
191
+
192
+ ## Citation
193
+ If you find this model useful in an academic context, please cite the technical report:
194
+ ```bibtex
195
+ @misc{zyphra2026zonos3mini,
196
+ title = {{ZONOS3-Mini Technical Report}},
197
+ author = {Gabriel Clark and Sofian Mejjoute and George Close and Beren Millidge},
198
+ year = {2026},
199
+ eprint = {2611.16525},
200
+ archivePrefix = {arXiv},
201
+ primaryClass = {cs.SD},
202
+ url = {https://arxiv.org/abs/2611.16525},
203
+ }
204
+ ```
assets/arch_zac1_dark.png ADDED

Git LFS Details

  • SHA256: 665c7e614d589e93379d134ad5177a3f4dabdc6e5e681a95f82908ae08321652
  • Pointer size: 131 Bytes
  • Size of remote file: 146 kB
assets/eval_pareto.png ADDED

Git LFS Details

  • SHA256: c225dcb8c2ff951b86705d6e96f1b5460f83aaf5eb2d104da8c4aefd71d703b5
  • Pointer size: 131 Bytes
  • Size of remote file: 309 kB
assets/zonos3_mini_arloop_animated.gif ADDED

Git LFS Details

  • SHA256: b81430e1ec1e1fc2383d877be8fa937fa0ec67efa24d75b3dacd11f009a57308
  • Pointer size: 132 Bytes
  • Size of remote file: 3.59 MB
assets/zonos3_mini_hero.webp ADDED

Git LFS Details

  • SHA256: 400b7056d52bb96f17ee65251a4127ad1ce1a84036a633f98325e57de693d6ba
  • Pointer size: 132 Bytes
  • Size of remote file: 7.01 MB
autoencoder/generator.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:75a587a3eda9ca3d345f0c305756a7ff8619b01685c261b08ae243c69c2caf9a
3
+ size 7960647220
autoencoder/heads/full_mr/config_full_unfreezeall_mr.json ADDED
@@ -0,0 +1,455 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": {
3
+ "architecture": "s1_semantic_rvq_moss_v2_hybrid",
4
+ "sample_rate": 44100,
5
+ "internal_sample_rate": 44160,
6
+ "channels": 1,
7
+ "channel_interleave": false,
8
+ "frame_rate": 15,
9
+ "internal_hop_length": 2944,
10
+ "external_samples_per_frame": 2940,
11
+ "encoder_stages": [
12
+ {
13
+ "type": "S1CausalWNConv1d",
14
+ "in_channels": 1,
15
+ "out_channels": 64,
16
+ "kernel_size": 7,
17
+ "stride": 1
18
+ },
19
+ {
20
+ "type": "S1EncoderBlock",
21
+ "in_channels": 64,
22
+ "out_channels": 128,
23
+ "stride": 2,
24
+ "kernel_size": 4,
25
+ "residual_dilations": [
26
+ 1,
27
+ 3,
28
+ 9
29
+ ],
30
+ "residual_kernel_size": 7,
31
+ "activation": "Snake1d",
32
+ "transformer_layers": 0
33
+ },
34
+ {
35
+ "type": "S1EncoderBlock",
36
+ "in_channels": 128,
37
+ "out_channels": 256,
38
+ "stride": 4,
39
+ "kernel_size": 8,
40
+ "residual_dilations": [
41
+ 1,
42
+ 3,
43
+ 9
44
+ ],
45
+ "residual_kernel_size": 7,
46
+ "activation": "Snake1d",
47
+ "transformer_layers": 0
48
+ },
49
+ {
50
+ "type": "PatchedPretransform",
51
+ "patch_size": 23,
52
+ "direction": "down"
53
+ },
54
+ {
55
+ "type": "MossProjectedTransformer",
56
+ "stage": 2,
57
+ "input_dimension": 5888,
58
+ "output_dimension": 384,
59
+ "d_model": 768,
60
+ "num_heads": 12,
61
+ "num_layers": 12,
62
+ "dim_feedforward": 3072,
63
+ "causal": true,
64
+ "norm": "layer_norm",
65
+ "positional_embedding": "rope",
66
+ "max_period": 10000,
67
+ "gating": "none",
68
+ "activation": "gelu",
69
+ "layer_scale": 0.01,
70
+ "attention_dropout": 0.1,
71
+ "ffn_dropout": 0.1,
72
+ "frame_rate": 240,
73
+ "context_duration": 2,
74
+ "context_frames": 480
75
+ },
76
+ {
77
+ "type": "PatchedPretransform",
78
+ "patch_size": 2,
79
+ "direction": "down"
80
+ },
81
+ {
82
+ "type": "MossProjectedTransformer",
83
+ "stage": 3,
84
+ "input_dimension": 768,
85
+ "output_dimension": 384,
86
+ "d_model": 768,
87
+ "num_heads": 12,
88
+ "num_layers": 12,
89
+ "dim_feedforward": 3072,
90
+ "causal": true,
91
+ "norm": "layer_norm",
92
+ "positional_embedding": "rope",
93
+ "max_period": 10000,
94
+ "gating": "none",
95
+ "activation": "gelu",
96
+ "layer_scale": 0.01,
97
+ "attention_dropout": 0.1,
98
+ "ffn_dropout": 0.1,
99
+ "frame_rate": 120,
100
+ "context_duration": 4,
101
+ "context_frames": 480
102
+ },
103
+ {
104
+ "type": "PatchedPretransform",
105
+ "patch_size": 2,
106
+ "direction": "down"
107
+ },
108
+ {
109
+ "type": "MossProjectedTransformer",
110
+ "stage": 4,
111
+ "input_dimension": 768,
112
+ "output_dimension": 384,
113
+ "d_model": 768,
114
+ "num_heads": 12,
115
+ "num_layers": 12,
116
+ "dim_feedforward": 3072,
117
+ "causal": true,
118
+ "norm": "layer_norm",
119
+ "positional_embedding": "rope",
120
+ "max_period": 10000,
121
+ "gating": "none",
122
+ "activation": "gelu",
123
+ "layer_scale": 0.01,
124
+ "attention_dropout": 0.1,
125
+ "ffn_dropout": 0.1,
126
+ "frame_rate": 60,
127
+ "context_duration": 8,
128
+ "context_frames": 480
129
+ },
130
+ {
131
+ "type": "PatchedPretransform",
132
+ "patch_size": 2,
133
+ "direction": "down"
134
+ },
135
+ {
136
+ "type": "MossProjectedTransformer",
137
+ "stage": 5,
138
+ "input_dimension": 768,
139
+ "output_dimension": 640,
140
+ "d_model": 768,
141
+ "num_heads": 12,
142
+ "num_layers": 12,
143
+ "dim_feedforward": 3072,
144
+ "causal": true,
145
+ "norm": "layer_norm",
146
+ "positional_embedding": "rope",
147
+ "max_period": 10000,
148
+ "gating": "none",
149
+ "activation": "gelu",
150
+ "layer_scale": 0.01,
151
+ "attention_dropout": 0.1,
152
+ "ffn_dropout": 0.1,
153
+ "frame_rate": 30,
154
+ "context_duration": 10,
155
+ "context_frames": 300
156
+ },
157
+ {
158
+ "type": "PatchedPretransform",
159
+ "patch_size": 2,
160
+ "direction": "down"
161
+ },
162
+ {
163
+ "type": "MossProjectedTransformer",
164
+ "stage": 6,
165
+ "input_dimension": 1280,
166
+ "output_dimension": 1024,
167
+ "d_model": 1280,
168
+ "num_heads": 20,
169
+ "num_layers": 32,
170
+ "dim_feedforward": 5120,
171
+ "causal": true,
172
+ "norm": "layer_norm",
173
+ "positional_embedding": "rope",
174
+ "max_period": 10000,
175
+ "gating": "none",
176
+ "activation": "gelu",
177
+ "layer_scale": 0.01,
178
+ "attention_dropout": 0.1,
179
+ "ffn_dropout": 0.1,
180
+ "frame_rate": 15,
181
+ "context_duration": 10,
182
+ "context_frames": 150
183
+ }
184
+ ],
185
+ "decoder_stages": [
186
+ {
187
+ "type": "MossProjectedTransformer",
188
+ "stage": 6,
189
+ "input_dimension": 1024,
190
+ "output_dimension": 1280,
191
+ "d_model": 1280,
192
+ "num_heads": 20,
193
+ "num_layers": 32,
194
+ "dim_feedforward": 5120,
195
+ "causal": true,
196
+ "norm": "layer_norm",
197
+ "positional_embedding": "rope",
198
+ "max_period": 10000,
199
+ "gating": "none",
200
+ "activation": "gelu",
201
+ "layer_scale": 0.01,
202
+ "attention_dropout": 0.1,
203
+ "ffn_dropout": 0.1,
204
+ "frame_rate": 15,
205
+ "context_duration": 10,
206
+ "context_frames": 150
207
+ },
208
+ {
209
+ "type": "PatchedPretransform",
210
+ "patch_size": 2,
211
+ "direction": "up"
212
+ },
213
+ {
214
+ "type": "MossProjectedTransformer",
215
+ "stage": 5,
216
+ "input_dimension": 640,
217
+ "output_dimension": 768,
218
+ "d_model": 768,
219
+ "num_heads": 12,
220
+ "num_layers": 12,
221
+ "dim_feedforward": 3072,
222
+ "causal": true,
223
+ "norm": "layer_norm",
224
+ "positional_embedding": "rope",
225
+ "max_period": 10000,
226
+ "gating": "none",
227
+ "activation": "gelu",
228
+ "layer_scale": 0.01,
229
+ "attention_dropout": 0.1,
230
+ "ffn_dropout": 0.1,
231
+ "frame_rate": 30,
232
+ "context_duration": 10,
233
+ "context_frames": 300
234
+ },
235
+ {
236
+ "type": "PatchedPretransform",
237
+ "patch_size": 2,
238
+ "direction": "up"
239
+ },
240
+ {
241
+ "type": "MossProjectedTransformer",
242
+ "stage": 4,
243
+ "input_dimension": 384,
244
+ "output_dimension": 768,
245
+ "d_model": 768,
246
+ "num_heads": 12,
247
+ "num_layers": 12,
248
+ "dim_feedforward": 3072,
249
+ "causal": true,
250
+ "norm": "layer_norm",
251
+ "positional_embedding": "rope",
252
+ "max_period": 10000,
253
+ "gating": "none",
254
+ "activation": "gelu",
255
+ "layer_scale": 0.01,
256
+ "attention_dropout": 0.1,
257
+ "ffn_dropout": 0.1,
258
+ "frame_rate": 60,
259
+ "context_duration": 8,
260
+ "context_frames": 480
261
+ },
262
+ {
263
+ "type": "PatchedPretransform",
264
+ "patch_size": 2,
265
+ "direction": "up"
266
+ },
267
+ {
268
+ "type": "MossProjectedTransformer",
269
+ "stage": 3,
270
+ "input_dimension": 384,
271
+ "output_dimension": 768,
272
+ "d_model": 768,
273
+ "num_heads": 12,
274
+ "num_layers": 12,
275
+ "dim_feedforward": 3072,
276
+ "causal": true,
277
+ "norm": "layer_norm",
278
+ "positional_embedding": "rope",
279
+ "max_period": 10000,
280
+ "gating": "none",
281
+ "activation": "gelu",
282
+ "layer_scale": 0.01,
283
+ "attention_dropout": 0.1,
284
+ "ffn_dropout": 0.1,
285
+ "frame_rate": 120,
286
+ "context_duration": 4,
287
+ "context_frames": 480
288
+ },
289
+ {
290
+ "type": "PatchedPretransform",
291
+ "patch_size": 2,
292
+ "direction": "up"
293
+ },
294
+ {
295
+ "type": "MossProjectedTransformer",
296
+ "stage": 2,
297
+ "input_dimension": 384,
298
+ "output_dimension": 5888,
299
+ "d_model": 768,
300
+ "num_heads": 12,
301
+ "num_layers": 12,
302
+ "dim_feedforward": 3072,
303
+ "causal": true,
304
+ "norm": "layer_norm",
305
+ "positional_embedding": "rope",
306
+ "max_period": 10000,
307
+ "gating": "none",
308
+ "activation": "gelu",
309
+ "layer_scale": 0.01,
310
+ "attention_dropout": 0.1,
311
+ "ffn_dropout": 0.1,
312
+ "frame_rate": 240,
313
+ "context_duration": 2,
314
+ "context_frames": 480
315
+ },
316
+ {
317
+ "type": "PatchedPretransform",
318
+ "patch_size": 23,
319
+ "direction": "up"
320
+ },
321
+ {
322
+ "type": "S1DecoderBlock",
323
+ "in_channels": 256,
324
+ "out_channels": 128,
325
+ "stride": 4,
326
+ "kernel_size": 8,
327
+ "residual_dilations": [
328
+ 1,
329
+ 3,
330
+ 9
331
+ ],
332
+ "residual_kernel_size": 7,
333
+ "activation": "Snake1d",
334
+ "transformer_layers": 0
335
+ },
336
+ {
337
+ "type": "S1DecoderBlock",
338
+ "in_channels": 128,
339
+ "out_channels": 64,
340
+ "stride": 2,
341
+ "kernel_size": 4,
342
+ "residual_dilations": [
343
+ 1,
344
+ 3,
345
+ 9
346
+ ],
347
+ "residual_kernel_size": 7,
348
+ "activation": "Snake1d",
349
+ "transformer_layers": 0
350
+ },
351
+ {
352
+ "type": "S1WaveformHead",
353
+ "in_channels": 64,
354
+ "out_channels": 1,
355
+ "kernel_size": 7,
356
+ "activation_before": "Snake1d",
357
+ "activation_after": "tanh"
358
+ }
359
+ ],
360
+ "bottleneck": {
361
+ "type": "dual_encoding_semantic_residual_rvq",
362
+ "input_dim": 1024,
363
+ "semantic_codebooks": 1,
364
+ "residual_codebooks": 9,
365
+ "semantic_codebook_size": 4096,
366
+ "residual_codebook_size": 4096,
367
+ "codebook_dim": 8,
368
+ "quantizer_dropout": 0.5,
369
+ "semantic_quantizer_dropout": 0.0,
370
+ "downsample_factors": [],
371
+ "upsample_factors": [],
372
+ "pre_module": "identity",
373
+ "post_module": "identity",
374
+ "lookup_and_vq_loss_dtype": "float32",
375
+ "semantic_head": {
376
+ "type": "ConvNeXtDecoder",
377
+ "width": 768,
378
+ "hidden": 2048,
379
+ "depth": 4,
380
+ "out_channels": 1024,
381
+ "causal": true
382
+ },
383
+ "distillation_input": "decoded_ssl_semantic_codebook"
384
+ },
385
+ "boundary_resampler": {
386
+ "input_rate": 44100,
387
+ "codec_rate": 44160,
388
+ "output_rate": 44100,
389
+ "encoder_ratio": [
390
+ 736,
391
+ 735
392
+ ],
393
+ "decoder_ratio": [
394
+ 735,
395
+ 736
396
+ ],
397
+ "method": "sinc_interp_hann",
398
+ "lowpass_filter_width": 16,
399
+ "rolloff": 0.99,
400
+ "streaming_policy": "finite_fir_lookahead_with_delay_compensated_timestamps",
401
+ "training_policy": "time_aligned_full_crop_reference",
402
+ "losses_and_teacher_use_external_rate": true
403
+ },
404
+ "checkpoint_transformers": true,
405
+ "attention_implementation": "pytorch_sdpa_chunked_causal_local_with_dropout",
406
+ "residual_codebooks": 9,
407
+ "semantic_codebook_size": 4096,
408
+ "residual_codebook_size": 4096,
409
+ "codebook_dim": 8,
410
+ "quantizer_dropout": 0.5,
411
+ "decoder_mirrors_encoder": true,
412
+ "dual_encoding": {
413
+ "width": 768,
414
+ "hidden": 2048,
415
+ "depth": 4,
416
+ "causal": true,
417
+ "window_ratio": 2.0,
418
+ "strategy": "frozen_ssl_to_q1_decoded_semantic_residual",
419
+ "normalization": "unchanged_baseline"
420
+ }
421
+ },
422
+ "head": {
423
+ "depth": 8,
424
+ "hidden": 2048,
425
+ "hop": 184,
426
+ "frame_rate": 240,
427
+ "trainable_stages": [
428
+ 0,
429
+ 1,
430
+ 2,
431
+ 3,
432
+ 4,
433
+ 5,
434
+ 6,
435
+ 7,
436
+ 8
437
+ ],
438
+ "parametrization": "per branch: log-magnitude + phase angle (Vocos), magnitude clipped at 1e2, periodic Hann, centered iSTFT; branches combined by a linear fp32 FIR mixer",
439
+ "recompute_trainable_stages": true,
440
+ "compile_trainable_stages": true,
441
+ "type": "mr_istft",
442
+ "n_ffts": [
443
+ 736,
444
+ 2208
445
+ ],
446
+ "fir_kernel": 127,
447
+ "warm_blocks": 4
448
+ },
449
+ "init": {
450
+ "generator_checkpoint": "hf://Zyphra/ZONOS3-Mini/autoencoder/generator.pt",
451
+ "trainable_rename": {
452
+ "head.out.": "head.outs.0."
453
+ }
454
+ }
455
+ }
autoencoder/heads/full_mr/step_0050000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d2e2340c74f763340535f84e5dda59b5eabea668360400f8fac6b641598a1e9
3
+ size 4012134189
autoencoder/heads/keep2_plain/config_full_unfreezeall_mr_drop30_4x10.json ADDED
@@ -0,0 +1,473 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model": {
3
+ "architecture": "s1_semantic_rvq_moss_v2_hybrid",
4
+ "sample_rate": 44100,
5
+ "internal_sample_rate": 44160,
6
+ "channels": 1,
7
+ "channel_interleave": false,
8
+ "frame_rate": 15,
9
+ "internal_hop_length": 2944,
10
+ "external_samples_per_frame": 2940,
11
+ "encoder_stages": [
12
+ {
13
+ "type": "S1CausalWNConv1d",
14
+ "in_channels": 1,
15
+ "out_channels": 64,
16
+ "kernel_size": 7,
17
+ "stride": 1
18
+ },
19
+ {
20
+ "type": "S1EncoderBlock",
21
+ "in_channels": 64,
22
+ "out_channels": 128,
23
+ "stride": 2,
24
+ "kernel_size": 4,
25
+ "residual_dilations": [
26
+ 1,
27
+ 3,
28
+ 9
29
+ ],
30
+ "residual_kernel_size": 7,
31
+ "activation": "Snake1d",
32
+ "transformer_layers": 0
33
+ },
34
+ {
35
+ "type": "S1EncoderBlock",
36
+ "in_channels": 128,
37
+ "out_channels": 256,
38
+ "stride": 4,
39
+ "kernel_size": 8,
40
+ "residual_dilations": [
41
+ 1,
42
+ 3,
43
+ 9
44
+ ],
45
+ "residual_kernel_size": 7,
46
+ "activation": "Snake1d",
47
+ "transformer_layers": 0
48
+ },
49
+ {
50
+ "type": "PatchedPretransform",
51
+ "patch_size": 23,
52
+ "direction": "down"
53
+ },
54
+ {
55
+ "type": "MossProjectedTransformer",
56
+ "stage": 2,
57
+ "input_dimension": 5888,
58
+ "output_dimension": 384,
59
+ "d_model": 768,
60
+ "num_heads": 12,
61
+ "num_layers": 12,
62
+ "dim_feedforward": 3072,
63
+ "causal": true,
64
+ "norm": "layer_norm",
65
+ "positional_embedding": "rope",
66
+ "max_period": 10000,
67
+ "gating": "none",
68
+ "activation": "gelu",
69
+ "layer_scale": 0.01,
70
+ "attention_dropout": 0.1,
71
+ "ffn_dropout": 0.1,
72
+ "frame_rate": 240,
73
+ "context_duration": 2,
74
+ "context_frames": 480
75
+ },
76
+ {
77
+ "type": "PatchedPretransform",
78
+ "patch_size": 2,
79
+ "direction": "down"
80
+ },
81
+ {
82
+ "type": "MossProjectedTransformer",
83
+ "stage": 3,
84
+ "input_dimension": 768,
85
+ "output_dimension": 384,
86
+ "d_model": 768,
87
+ "num_heads": 12,
88
+ "num_layers": 12,
89
+ "dim_feedforward": 3072,
90
+ "causal": true,
91
+ "norm": "layer_norm",
92
+ "positional_embedding": "rope",
93
+ "max_period": 10000,
94
+ "gating": "none",
95
+ "activation": "gelu",
96
+ "layer_scale": 0.01,
97
+ "attention_dropout": 0.1,
98
+ "ffn_dropout": 0.1,
99
+ "frame_rate": 120,
100
+ "context_duration": 4,
101
+ "context_frames": 480
102
+ },
103
+ {
104
+ "type": "PatchedPretransform",
105
+ "patch_size": 2,
106
+ "direction": "down"
107
+ },
108
+ {
109
+ "type": "MossProjectedTransformer",
110
+ "stage": 4,
111
+ "input_dimension": 768,
112
+ "output_dimension": 384,
113
+ "d_model": 768,
114
+ "num_heads": 12,
115
+ "num_layers": 12,
116
+ "dim_feedforward": 3072,
117
+ "causal": true,
118
+ "norm": "layer_norm",
119
+ "positional_embedding": "rope",
120
+ "max_period": 10000,
121
+ "gating": "none",
122
+ "activation": "gelu",
123
+ "layer_scale": 0.01,
124
+ "attention_dropout": 0.1,
125
+ "ffn_dropout": 0.1,
126
+ "frame_rate": 60,
127
+ "context_duration": 8,
128
+ "context_frames": 480
129
+ },
130
+ {
131
+ "type": "PatchedPretransform",
132
+ "patch_size": 2,
133
+ "direction": "down"
134
+ },
135
+ {
136
+ "type": "MossProjectedTransformer",
137
+ "stage": 5,
138
+ "input_dimension": 768,
139
+ "output_dimension": 640,
140
+ "d_model": 768,
141
+ "num_heads": 12,
142
+ "num_layers": 12,
143
+ "dim_feedforward": 3072,
144
+ "causal": true,
145
+ "norm": "layer_norm",
146
+ "positional_embedding": "rope",
147
+ "max_period": 10000,
148
+ "gating": "none",
149
+ "activation": "gelu",
150
+ "layer_scale": 0.01,
151
+ "attention_dropout": 0.1,
152
+ "ffn_dropout": 0.1,
153
+ "frame_rate": 30,
154
+ "context_duration": 10,
155
+ "context_frames": 300
156
+ },
157
+ {
158
+ "type": "PatchedPretransform",
159
+ "patch_size": 2,
160
+ "direction": "down"
161
+ },
162
+ {
163
+ "type": "MossProjectedTransformer",
164
+ "stage": 6,
165
+ "input_dimension": 1280,
166
+ "output_dimension": 1024,
167
+ "d_model": 1280,
168
+ "num_heads": 20,
169
+ "num_layers": 32,
170
+ "dim_feedforward": 5120,
171
+ "causal": true,
172
+ "norm": "layer_norm",
173
+ "positional_embedding": "rope",
174
+ "max_period": 10000,
175
+ "gating": "none",
176
+ "activation": "gelu",
177
+ "layer_scale": 0.01,
178
+ "attention_dropout": 0.1,
179
+ "ffn_dropout": 0.1,
180
+ "frame_rate": 15,
181
+ "context_duration": 10,
182
+ "context_frames": 150
183
+ }
184
+ ],
185
+ "decoder_stages": [
186
+ {
187
+ "type": "MossProjectedTransformer",
188
+ "stage": 6,
189
+ "input_dimension": 1024,
190
+ "output_dimension": 1280,
191
+ "d_model": 1280,
192
+ "num_heads": 20,
193
+ "num_layers": 32,
194
+ "dim_feedforward": 5120,
195
+ "causal": true,
196
+ "norm": "layer_norm",
197
+ "positional_embedding": "rope",
198
+ "max_period": 10000,
199
+ "gating": "none",
200
+ "activation": "gelu",
201
+ "layer_scale": 0.01,
202
+ "attention_dropout": 0.1,
203
+ "ffn_dropout": 0.1,
204
+ "frame_rate": 15,
205
+ "context_duration": 10,
206
+ "context_frames": 150
207
+ },
208
+ {
209
+ "type": "PatchedPretransform",
210
+ "patch_size": 2,
211
+ "direction": "up"
212
+ },
213
+ {
214
+ "type": "MossProjectedTransformer",
215
+ "stage": 5,
216
+ "input_dimension": 640,
217
+ "output_dimension": 768,
218
+ "d_model": 768,
219
+ "num_heads": 12,
220
+ "num_layers": 12,
221
+ "dim_feedforward": 3072,
222
+ "causal": true,
223
+ "norm": "layer_norm",
224
+ "positional_embedding": "rope",
225
+ "max_period": 10000,
226
+ "gating": "none",
227
+ "activation": "gelu",
228
+ "layer_scale": 0.01,
229
+ "attention_dropout": 0.1,
230
+ "ffn_dropout": 0.1,
231
+ "frame_rate": 30,
232
+ "context_duration": 10,
233
+ "context_frames": 300
234
+ },
235
+ {
236
+ "type": "PatchedPretransform",
237
+ "patch_size": 2,
238
+ "direction": "up"
239
+ },
240
+ {
241
+ "type": "MossProjectedTransformer",
242
+ "stage": 4,
243
+ "input_dimension": 384,
244
+ "output_dimension": 768,
245
+ "d_model": 768,
246
+ "num_heads": 12,
247
+ "num_layers": 12,
248
+ "dim_feedforward": 3072,
249
+ "causal": true,
250
+ "norm": "layer_norm",
251
+ "positional_embedding": "rope",
252
+ "max_period": 10000,
253
+ "gating": "none",
254
+ "activation": "gelu",
255
+ "layer_scale": 0.01,
256
+ "attention_dropout": 0.1,
257
+ "ffn_dropout": 0.1,
258
+ "frame_rate": 60,
259
+ "context_duration": 8,
260
+ "context_frames": 480
261
+ },
262
+ {
263
+ "type": "PatchedPretransform",
264
+ "patch_size": 2,
265
+ "direction": "up"
266
+ },
267
+ {
268
+ "type": "MossProjectedTransformer",
269
+ "stage": 3,
270
+ "input_dimension": 384,
271
+ "output_dimension": 768,
272
+ "d_model": 768,
273
+ "num_heads": 12,
274
+ "num_layers": 12,
275
+ "dim_feedforward": 3072,
276
+ "causal": true,
277
+ "norm": "layer_norm",
278
+ "positional_embedding": "rope",
279
+ "max_period": 10000,
280
+ "gating": "none",
281
+ "activation": "gelu",
282
+ "layer_scale": 0.01,
283
+ "attention_dropout": 0.1,
284
+ "ffn_dropout": 0.1,
285
+ "frame_rate": 120,
286
+ "context_duration": 4,
287
+ "context_frames": 480
288
+ },
289
+ {
290
+ "type": "PatchedPretransform",
291
+ "patch_size": 2,
292
+ "direction": "up"
293
+ },
294
+ {
295
+ "type": "MossProjectedTransformer",
296
+ "stage": 2,
297
+ "input_dimension": 384,
298
+ "output_dimension": 5888,
299
+ "d_model": 768,
300
+ "num_heads": 12,
301
+ "num_layers": 12,
302
+ "dim_feedforward": 3072,
303
+ "causal": true,
304
+ "norm": "layer_norm",
305
+ "positional_embedding": "rope",
306
+ "max_period": 10000,
307
+ "gating": "none",
308
+ "activation": "gelu",
309
+ "layer_scale": 0.01,
310
+ "attention_dropout": 0.1,
311
+ "ffn_dropout": 0.1,
312
+ "frame_rate": 240,
313
+ "context_duration": 2,
314
+ "context_frames": 480
315
+ },
316
+ {
317
+ "type": "PatchedPretransform",
318
+ "patch_size": 23,
319
+ "direction": "up"
320
+ },
321
+ {
322
+ "type": "S1DecoderBlock",
323
+ "in_channels": 256,
324
+ "out_channels": 128,
325
+ "stride": 4,
326
+ "kernel_size": 8,
327
+ "residual_dilations": [
328
+ 1,
329
+ 3,
330
+ 9
331
+ ],
332
+ "residual_kernel_size": 7,
333
+ "activation": "Snake1d",
334
+ "transformer_layers": 0
335
+ },
336
+ {
337
+ "type": "S1DecoderBlock",
338
+ "in_channels": 128,
339
+ "out_channels": 64,
340
+ "stride": 2,
341
+ "kernel_size": 4,
342
+ "residual_dilations": [
343
+ 1,
344
+ 3,
345
+ 9
346
+ ],
347
+ "residual_kernel_size": 7,
348
+ "activation": "Snake1d",
349
+ "transformer_layers": 0
350
+ },
351
+ {
352
+ "type": "S1WaveformHead",
353
+ "in_channels": 64,
354
+ "out_channels": 1,
355
+ "kernel_size": 7,
356
+ "activation_before": "Snake1d",
357
+ "activation_after": "tanh"
358
+ }
359
+ ],
360
+ "bottleneck": {
361
+ "type": "dual_encoding_semantic_residual_rvq",
362
+ "input_dim": 1024,
363
+ "semantic_codebooks": 1,
364
+ "residual_codebooks": 9,
365
+ "semantic_codebook_size": 4096,
366
+ "residual_codebook_size": 4096,
367
+ "codebook_dim": 8,
368
+ "quantizer_dropout": 0.5,
369
+ "semantic_quantizer_dropout": 0.0,
370
+ "downsample_factors": [],
371
+ "upsample_factors": [],
372
+ "pre_module": "identity",
373
+ "post_module": "identity",
374
+ "lookup_and_vq_loss_dtype": "float32",
375
+ "semantic_head": {
376
+ "type": "ConvNeXtDecoder",
377
+ "width": 768,
378
+ "hidden": 2048,
379
+ "depth": 4,
380
+ "out_channels": 1024,
381
+ "causal": true
382
+ },
383
+ "distillation_input": "decoded_ssl_semantic_codebook"
384
+ },
385
+ "boundary_resampler": {
386
+ "input_rate": 44100,
387
+ "codec_rate": 44160,
388
+ "output_rate": 44100,
389
+ "encoder_ratio": [
390
+ 736,
391
+ 735
392
+ ],
393
+ "decoder_ratio": [
394
+ 735,
395
+ 736
396
+ ],
397
+ "method": "sinc_interp_hann",
398
+ "lowpass_filter_width": 16,
399
+ "rolloff": 0.99,
400
+ "streaming_policy": "finite_fir_lookahead_with_delay_compensated_timestamps",
401
+ "training_policy": "time_aligned_full_crop_reference",
402
+ "losses_and_teacher_use_external_rate": true
403
+ },
404
+ "checkpoint_transformers": true,
405
+ "attention_implementation": "pytorch_sdpa_chunked_causal_local_with_dropout",
406
+ "residual_codebooks": 9,
407
+ "semantic_codebook_size": 4096,
408
+ "residual_codebook_size": 4096,
409
+ "codebook_dim": 8,
410
+ "quantizer_dropout": 0.5,
411
+ "decoder_mirrors_encoder": true,
412
+ "dual_encoding": {
413
+ "width": 768,
414
+ "hidden": 2048,
415
+ "depth": 4,
416
+ "causal": true,
417
+ "window_ratio": 2.0,
418
+ "strategy": "frozen_ssl_to_q1_decoded_semantic_residual",
419
+ "normalization": "unchanged_baseline"
420
+ }
421
+ },
422
+ "head": {
423
+ "depth": 8,
424
+ "hidden": 2048,
425
+ "hop": 184,
426
+ "frame_rate": 240,
427
+ "trainable_stages": [
428
+ 0,
429
+ 1,
430
+ 2,
431
+ 3,
432
+ 4,
433
+ 5,
434
+ 6,
435
+ 7,
436
+ 8
437
+ ],
438
+ "parametrization": "per branch: log-magnitude + phase angle (Vocos), magnitude clipped at 1e2, periodic Hann, centered iSTFT; branches combined by a linear fp32 FIR mixer",
439
+ "recompute_trainable_stages": true,
440
+ "compile_trainable_stages": true,
441
+ "type": "mr_istft",
442
+ "n_ffts": [
443
+ 736,
444
+ 2208
445
+ ],
446
+ "fir_kernel": 127,
447
+ "stage_layers": {
448
+ "0": [
449
+ 0,
450
+ 31
451
+ ],
452
+ "2": [
453
+ 0,
454
+ 11
455
+ ],
456
+ "4": [
457
+ 0,
458
+ 11
459
+ ],
460
+ "6": [
461
+ 0,
462
+ 11
463
+ ],
464
+ "8": [
465
+ 0,
466
+ 11
467
+ ]
468
+ }
469
+ },
470
+ "init": {
471
+ "generator_checkpoint": "hf://Zyphra/ZONOS3-Mini/autoencoder/generator.pt"
472
+ }
473
+ }
autoencoder/heads/keep2_plain/step_0050000.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7432e335e717417961267f2242547ce389a796cdc296a8b8d5289658e3c28936
3
+ size 518498361
autoencoder/teacher/config.json ADDED
@@ -0,0 +1,81 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.0,
3
+ "adapter_act": "relu",
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": false,
8
+ "architectures": [
9
+ "Wav2Vec2BertModel"
10
+ ],
11
+ "attention_dropout": 0.0,
12
+ "bos_token_id": 1,
13
+ "classifier_proj_size": 768,
14
+ "codevector_dim": 768,
15
+ "conformer_conv_dropout": 0.1,
16
+ "contrastive_logits_temperature": 0.1,
17
+ "conv_depthwise_kernel_size": 31,
18
+ "ctc_loss_reduction": "sum",
19
+ "ctc_zero_infinity": false,
20
+ "diversity_loss_weight": 0.1,
21
+ "dtype": "bfloat16",
22
+ "eos_token_id": 2,
23
+ "feat_proj_dropout": 0.0,
24
+ "feat_quantizer_dropout": 0.0,
25
+ "feature_projection_input_dim": 160,
26
+ "final_dropout": 0.1,
27
+ "hidden_act": "swish",
28
+ "hidden_dropout": 0.0,
29
+ "hidden_size": 1024,
30
+ "initializer_range": 0.02,
31
+ "intermediate_size": 4096,
32
+ "layer_norm_eps": 1e-05,
33
+ "layerdrop": 0.1,
34
+ "left_max_position_embeddings": 64,
35
+ "mask_feature_length": 10,
36
+ "mask_feature_min_masks": 0,
37
+ "mask_feature_prob": 0.0,
38
+ "mask_time_length": 10,
39
+ "mask_time_min_masks": 2,
40
+ "mask_time_prob": 0.05,
41
+ "max_source_positions": 5000,
42
+ "model_type": "wav2vec2-bert",
43
+ "num_adapter_layers": 1,
44
+ "num_attention_heads": 16,
45
+ "num_codevector_groups": 2,
46
+ "num_codevectors_per_group": 320,
47
+ "num_hidden_layers": 16,
48
+ "num_negatives": 100,
49
+ "output_hidden_size": 1024,
50
+ "pad_token_id": 0,
51
+ "position_embeddings_type": "relative_key",
52
+ "proj_codevector_dim": 768,
53
+ "right_max_position_embeddings": 8,
54
+ "rotary_embedding_base": 10000,
55
+ "tdnn_dilation": [
56
+ 1,
57
+ 2,
58
+ 3,
59
+ 1,
60
+ 1
61
+ ],
62
+ "tdnn_dim": [
63
+ 512,
64
+ 512,
65
+ 512,
66
+ 512,
67
+ 1500
68
+ ],
69
+ "tdnn_kernel": [
70
+ 5,
71
+ 3,
72
+ 3,
73
+ 1,
74
+ 1
75
+ ],
76
+ "transformers_version": "5.3.0",
77
+ "use_intermediate_ffn_before_adapter": false,
78
+ "use_weighted_layer_sum": false,
79
+ "vocab_size": null,
80
+ "xvector_output_dim": 512
81
+ }
autoencoder/teacher/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a15182741d6ee3cd780b8b830f49fa6b7fcce383655d451970841b24114767a5
3
+ size 774162368
gguf/backbone-ptq-edit.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7e23d261e2cef3cbbbaa33f43fbe683acb393a157ac672f2de04858aef21011f
3
+ size 663491840
gguf/backbone-ptq.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66fceb6d73a8c30bde80249e4112bb308380ab7f3384fa11c8e4048115e4ceec
3
+ size 663150624
gguf/edit/accents.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac77fc95e91a6b0c28135f27ee3e4fef1411b8ac3dba969e06f99ce9f84e7376
3
+ size 256592064
gguf/spk-encoder.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9539f340b70be246005403b9c254732e180e85e0b3c57fb1973f0d6bc6c59bdc
3
+ size 45385728
gguf/ssl-teacher.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59e4f150e1ccdf54e6349b06605c2274f2262139678688025b6614f1393e415a
3
+ size 1548700000
gguf/steer.gguf ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e8e3f86e89a09801ce083b9286f43263ec696acfaff72e25aca537aa9efa90c
3
+ size 204960
gguf/voices/English/Cillian_en-Irish_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dadae1124c63cda7a0627fd282d54aa255a8800155eae1b9da54b08b345c2f10
3
+ size 5928
gguf/voices/English/Cillian_en-Irish_M_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f3125fae2f679fb5c8128d308ddd02986fd7d914975d17cc6dadf9eb130b1061
3
+ size 896
gguf/voices/English/Cillian_en-Irish_M_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d43eda64c4928dfb30ecf547e4a00100b8f83ed0eb5c1e07116725794da72ea2
3
+ size 852644
gguf/voices/English/Imogen_en-British_F_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c906f6c405c9078607fe9c54cf1b71ef0e112839490499e34bf8791ed1ad7e86
3
+ size 7728
gguf/voices/English/Imogen_en-British_F_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:683424b91faeade48c8a92e05e1c3679366abbba443ef62eccba8dc40de48e9e
3
+ size 896
gguf/voices/English/Imogen_en-British_F_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dde4691a6ccd95f4928fbbc4e2d49089824b264575bc0bf0ac4986dd2f8082c5
3
+ size 1117244
gguf/voices/English/Jett_en-energetic_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8f32e946509673ae422c37cc9b79957cb569f212360745f9110d343f6f1f9f2c
3
+ size 6608
gguf/voices/English/Jett_en-energetic_M_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:624b749d220dac7fdcda479b1964cb8f7de58c838c1357f2f0161dac06e34f48
3
+ size 896
gguf/voices/English/Jett_en-energetic_M_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:107c8350f191d47401da6844158528d0f493cbf46be4dee6bb81dd71bc052619
3
+ size 952604
gguf/voices/English/Morgana_en-witch_F_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2f60227a9f7cdc6ce696548f4c28c199d3f5afeadf7317a1bde132575a1ed731
3
+ size 12448
gguf/voices/English/Morgana_en-witch_F_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a817387697c43138639aa1e52a9e42d9f3a629655e0faea5ac8565eb8e57c091
3
+ size 896
gguf/voices/English/Morgana_en-witch_F_1.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8fec617ca7a4044f930c108ded0daf814ca93e61d97a7df626373aa182240ece
3
+ size 329812
gguf/voices/English/Rupert_en-British_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:abf42171c7402fea5888ec350a50b64fd78777e44d7caf665c04e4e2b629fc79
3
+ size 8048
gguf/voices/English/Rupert_en-British_M_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c68e699376dbafdc1ab0fadc133f48e7df9cb33b3a7b80b05f42c5c0aad8fa7
3
+ size 896
gguf/voices/English/Rupert_en-British_M_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3743b94cfc35c004fb5c2bcc9c2a537b605e3e15d4677e2b5b811bc18d95151e
3
+ size 1164284
gguf/voices/English/Sakura_en-anime_F_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4b8c8288b45e1a3ff5aa3adc05a4af386f5889e9ab47902ff36f9daa8900ded2
3
+ size 9168
gguf/voices/English/Sakura_en-anime_F_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:daf3e82d2e6f2ce2fe90c73473a312a721295ae7f85763f99c87ee2f861063a9
3
+ size 896
gguf/voices/English/Sakura_en-anime_F_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:01e8f47241546efd1223bd1cac09ea81f1d9ae2cf3148260f37413f05db40229
3
+ size 1328924
gguf/voices/English/Savannah_en-American_F_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6e450fea8fa521c6be32b64fc61f48d6f9fd8301fe5759abd2315ce6ac946fb
3
+ size 6968
gguf/voices/English/Savannah_en-American_F_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dd81587e66f96fb62da47ef13d5d867de464ca386a83f01b4403cdd19629ca6c
3
+ size 896
gguf/voices/English/Savannah_en-American_F_1.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8d5eaf8f6c18a0a3b452652da62d821f819d4e95f6347aa29986a503e219d14
3
+ size 183527
gguf/voices/English/Thaddeus_en-announcervoice_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c76de5bf16fd10282fced313c6c00a03fa2150b1df4c2382e19c6cbad65f179
3
+ size 9368
gguf/voices/English/Thaddeus_en-announcervoice_M_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fc8e21356730988aa20da70123991404345e0bdbfe3afe31e6371d8d93095461
3
+ size 896
gguf/voices/English/Thaddeus_en-announcervoice_M_1.mp3 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7fc96a18645b129b1e55659379ba796f5cfd70a800068b977d37f1254e93c52
3
+ size 247474
gguf/voices/English/Willow_en-asrm_F_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29ca4c14c16bb09b2bc48bb93f2d5a56327558d64271e54ebd1f713a4204905b
3
+ size 15048
gguf/voices/English/Willow_en-asrm_F_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ebb84309f21f1dcbe6371e2355dc98f8ee3d0aee5850c305d06a73fe0a7c222
3
+ size 896
gguf/voices/English/Willow_en-asrm_F_1.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6a7574e4b91794390867027eb5c2c2a12f6cef557fc7d804b45f985144a50e5e
3
+ size 2193284
gguf/voices/English/Wyatt_en-American_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a854f9b66eefa9203a9f752c2a1ec808345b4d72ac34f8547ee776c58e4b72f
3
+ size 10248
gguf/voices/English/Wyatt_en-American_M_1.embedding.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b39d377b653ebaca6bbd4ecd13c382fe16d81f0cfae27217d9b626e5589a0511
3
+ size 896
gguf/voices/English/Wyatt_en-American_M_1.ogg ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92c17f6eff4c9612ece96f0efae6f74f8e6b381f8c851ce739cdf0854a5a5831
3
+ size 147400
gguf/voices/French/Gaspard_fr_M_1.codes.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34293ea302d3a47c17ab2f8514fb3a3354dc4ad695581d9b2f93fb37476ee1f7
3
+ size 11488