Ref2VA/audio_vae/config.json
1.9 KB · 82 lines · json Raw
1 {
2 "_class_name": "MiniMaxH3AudioVAE",
3 "_diffusers_version": "0.32.2",
4 "auto_map": {
5 "AutoModel": "minimax_h3_audio_vae.MiniMaxH3AudioVAE"
6 },
7 "output_channel": 2,
8 "sample_rate": 32000,
9 "source_config_path": "config.yaml",
10 "source_safetensors_path": "model.safetensors",
11 "source_metadata_path": "metadata.json",
12 "latent_channels": 32,
13 "latents_mean": [
14 -0.020211687488382354,
15 0.3876466479950502,
16 -0.04398279799186767,
17 -0.28591514936373,
18 0.08179686214561671,
19 -0.35782641352446604,
20 0.040623809960919084,
21 -0.01552534501956604,
22 -0.223362481667332,
23 0.1821006842509091,
24 0.2941778783780663,
25 -0.07901167601970885,
26 -0.056815072777201,
27 -0.3699028221860095,
28 -0.31616315591624855,
29 0.5905951377425391,
30 -0.052139568068853864,
31 0.013673160263486295,
32 -0.03691647864630577,
33 0.09732660653298163,
34 -0.3394662328788498,
35 -0.30685677538541667,
36 -0.24504598907458763,
37 -0.034698524462007344,
38 0.02868032184767538,
39 -0.21217779266454084,
40 -0.1678263169941987,
41 0.3221287889040614,
42 -0.1223055851554907,
43 0.4356604928128464,
44 -0.0502599202236253,
45 0.3979258376211797
46 ],
47 "latents_std": [
48 1.6895524230479284,
49 2.76263727217653,
50 1.7945344281264435,
51 1.6801681847309828,
52 1.6390226546605453,
53 2.7788298348882177,
54 1.7659090095747236,
55 1.6199757612137327,
56 2.6336525640336896,
57 1.8539356672817833,
58 2.5056497896915633,
59 1.811019237886178,
60 1.9579657790720237,
61 1.6685498243529284,
62 1.4922469314453364,
63 3.298670198067373,
64 1.9491804496832168,
65 1.8720003270431442,
66 1.8334080103291832,
67 1.6488070416529093,
68 1.6176957696319716,
69 1.9131449234774398,
70 1.5695245398428617,
71 1.6943659940415912,
72 1.8318420762504692,
73 1.5540637421583379,
74 1.9344930328968526,
75 1.599198216109855,
76 1.718045989838149,
77 1.6307219190837705,
78 1.8661226051202384,
79 1.5613768203168363
80 ]
81 }
82