Hugging Face's logo

tky823
/
MusicFMInference

Safetensors
musicfm_inference
musicfm
custom_code
Model card Files
xet
Community
tky823 commited on
Commit
610b667
·
verified ·
1 Parent(s): f80fc53

Upload config.json with huggingface_hub

Browse files
Files changed (1) hide show
  1. config.json +92 -0
config.json ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "MusicFM25HzInference"
4
+ ],
5
+ "auto_map": {
6
+ "AutoConfig": "configuration_musicfm.MusicFMInferenceConfig",
7
+ "AutoModel": "modeling_musicfm.MusicFM25HzInference"
8
+ },
9
+ "codebook_dim": 16,
10
+ "codebook_size": 4096,
11
+ "conv_dim": 512,
12
+ "encoder_depth": 12,
13
+ "encoder_dim": 1024,
14
+ "features": [
15
+ "melspec_2048"
16
+ ],
17
+ "hop_length": 240,
18
+ "is_flash": false,
19
+ "layer_index": 9,
20
+ "mask_hop": 0.4,
21
+ "mask_prob": 0.6,
22
+ "model_type": "musicfm_inference",
23
+ "n_mels": 128,
24
+ "num_codebooks": 1,
25
+ "stat": {
26
+ "chromagram_1024_cnt": 1339008768,
27
+ "chromagram_1024_mean": 589.2947481634608,
28
+ "chromagram_1024_std": 913.857929063196,
29
+ "chromagram_2048_cnt": 1339008768,
30
+ "chromagram_2048_mean": 2062.286388327397,
31
+ "chromagram_2048_std": 3458.92657915397,
32
+ "chromagram_256_cnt": 1339008768,
33
+ "chromagram_256_mean": 55.15596556703181,
34
+ "chromagram_256_std": 73.91858278719991,
35
+ "chromagram_4096_cnt": 1339008768,
36
+ "chromagram_4096_mean": 7673.039107997085,
37
+ "chromagram_4096_std": 13009.883158267234,
38
+ "chromagram_512_cnt": 1339008768,
39
+ "chromagram_512_mean": 175.73092252759895,
40
+ "chromagram_512_std": 248.48485148525953,
41
+ "cqt_cnt": 9373061376,
42
+ "cqt_mean": 0.46341379757927165,
43
+ "cqt_std": 0.9543998080910191,
44
+ "melspec_1024_cnt": 14282760192,
45
+ "melspec_1024_mean": 0.37302579246531126,
46
+ "melspec_1024_std": 18.684082325919388,
47
+ "melspec_2048_cnt": 14282760192,
48
+ "melspec_2048_mean": 6.768444971712967,
49
+ "melspec_2048_std": 18.417922652295623,
50
+ "melspec_256_cnt": 14282760192,
51
+ "melspec_256_mean": -26.962600400166156,
52
+ "melspec_256_std": 36.13614100912126,
53
+ "melspec_4096_cnt": 14282760192,
54
+ "melspec_4096_mean": 13.617164614990036,
55
+ "melspec_4096_std": 18.08552130124525,
56
+ "melspec_512_cnt": 14282760192,
57
+ "melspec_512_mean": -9.108344167718862,
58
+ "melspec_512_std": 24.71910937988429,
59
+ "mfcc_1024_cnt": 1339008768,
60
+ "mfcc_1024_mean": 6.606636263169779,
61
+ "mfcc_1024_std": 34.151644801729624,
62
+ "mfcc_2048_cnt": 1339008768,
63
+ "mfcc_2048_mean": 5.281600844245184,
64
+ "mfcc_2048_std": 33.12784541220003,
65
+ "mfcc_256_cnt": 1339008768,
66
+ "mfcc_256_mean": -11.681755459447485,
67
+ "mfcc_256_std": 29.183186444668316,
68
+ "mfcc_4096_cnt": 1339008768,
69
+ "mfcc_4096_mean": 4.7616569480166095,
70
+ "mfcc_4096_std": 32.61458906894133,
71
+ "mfcc_512_cnt": 1339008768,
72
+ "mfcc_512_mean": -2.540581461792183,
73
+ "mfcc_512_std": 31.93752185832081,
74
+ "spec_1024_cnt": 57242624832,
75
+ "spec_1024_mean": -18.852271129208273,
76
+ "spec_1024_std": 26.443154583585663,
77
+ "spec_2048_cnt": 114373665600,
78
+ "spec_2048_mean": -15.638743433896792,
79
+ "spec_2048_std": 26.115825961611545,
80
+ "spec_256_cnt": 14394344256,
81
+ "spec_256_mean": -23.34296658431829,
82
+ "spec_256_std": 26.189295587132637,
83
+ "spec_4096_cnt": 228635747136,
84
+ "spec_4096_mean": -11.715532502794836,
85
+ "spec_4096_std": 25.763972210234062,
86
+ "spec_512_cnt": 28677104448,
87
+ "spec_512_mean": -21.31267396860235,
88
+ "spec_512_std": 26.52644536245769
89
+ },
90
+ "torch_dtype": "float32",
91
+ "transformers_version": "4.53.0"
92
+ }