Upload config.json with huggingface_hub
Browse files- config.json +91 -0
config.json
ADDED
|
@@ -0,0 +1,91 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"MusicFM25Hz"
|
| 4 |
+
],
|
| 5 |
+
"auto_map": {
|
| 6 |
+
"AutoConfig": "configuration_musicfm.MusicFMConfig",
|
| 7 |
+
"AutoModel": "modeling_musicfm.MusicFM25Hz"
|
| 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 |
+
"mask_hop": 0.4,
|
| 20 |
+
"mask_prob": 0.6,
|
| 21 |
+
"model_type": "musicfm",
|
| 22 |
+
"n_mels": 128,
|
| 23 |
+
"num_codebooks": 1,
|
| 24 |
+
"stat": {
|
| 25 |
+
"chromagram_1024_cnt": 1339008768,
|
| 26 |
+
"chromagram_1024_mean": 589.2947481634608,
|
| 27 |
+
"chromagram_1024_std": 913.857929063196,
|
| 28 |
+
"chromagram_2048_cnt": 1339008768,
|
| 29 |
+
"chromagram_2048_mean": 2062.286388327397,
|
| 30 |
+
"chromagram_2048_std": 3458.92657915397,
|
| 31 |
+
"chromagram_256_cnt": 1339008768,
|
| 32 |
+
"chromagram_256_mean": 55.15596556703181,
|
| 33 |
+
"chromagram_256_std": 73.91858278719991,
|
| 34 |
+
"chromagram_4096_cnt": 1339008768,
|
| 35 |
+
"chromagram_4096_mean": 7673.039107997085,
|
| 36 |
+
"chromagram_4096_std": 13009.883158267234,
|
| 37 |
+
"chromagram_512_cnt": 1339008768,
|
| 38 |
+
"chromagram_512_mean": 175.73092252759895,
|
| 39 |
+
"chromagram_512_std": 248.48485148525953,
|
| 40 |
+
"cqt_cnt": 9373061376,
|
| 41 |
+
"cqt_mean": 0.46341379757927165,
|
| 42 |
+
"cqt_std": 0.9543998080910191,
|
| 43 |
+
"melspec_1024_cnt": 14282760192,
|
| 44 |
+
"melspec_1024_mean": 0.37302579246531126,
|
| 45 |
+
"melspec_1024_std": 18.684082325919388,
|
| 46 |
+
"melspec_2048_cnt": 14282760192,
|
| 47 |
+
"melspec_2048_mean": 6.768444971712967,
|
| 48 |
+
"melspec_2048_std": 18.417922652295623,
|
| 49 |
+
"melspec_256_cnt": 14282760192,
|
| 50 |
+
"melspec_256_mean": -26.962600400166156,
|
| 51 |
+
"melspec_256_std": 36.13614100912126,
|
| 52 |
+
"melspec_4096_cnt": 14282760192,
|
| 53 |
+
"melspec_4096_mean": 13.617164614990036,
|
| 54 |
+
"melspec_4096_std": 18.08552130124525,
|
| 55 |
+
"melspec_512_cnt": 14282760192,
|
| 56 |
+
"melspec_512_mean": -9.108344167718862,
|
| 57 |
+
"melspec_512_std": 24.71910937988429,
|
| 58 |
+
"mfcc_1024_cnt": 1339008768,
|
| 59 |
+
"mfcc_1024_mean": 6.606636263169779,
|
| 60 |
+
"mfcc_1024_std": 34.151644801729624,
|
| 61 |
+
"mfcc_2048_cnt": 1339008768,
|
| 62 |
+
"mfcc_2048_mean": 5.281600844245184,
|
| 63 |
+
"mfcc_2048_std": 33.12784541220003,
|
| 64 |
+
"mfcc_256_cnt": 1339008768,
|
| 65 |
+
"mfcc_256_mean": -11.681755459447485,
|
| 66 |
+
"mfcc_256_std": 29.183186444668316,
|
| 67 |
+
"mfcc_4096_cnt": 1339008768,
|
| 68 |
+
"mfcc_4096_mean": 4.7616569480166095,
|
| 69 |
+
"mfcc_4096_std": 32.61458906894133,
|
| 70 |
+
"mfcc_512_cnt": 1339008768,
|
| 71 |
+
"mfcc_512_mean": -2.540581461792183,
|
| 72 |
+
"mfcc_512_std": 31.93752185832081,
|
| 73 |
+
"spec_1024_cnt": 57242624832,
|
| 74 |
+
"spec_1024_mean": -18.852271129208273,
|
| 75 |
+
"spec_1024_std": 26.443154583585663,
|
| 76 |
+
"spec_2048_cnt": 114373665600,
|
| 77 |
+
"spec_2048_mean": -15.638743433896792,
|
| 78 |
+
"spec_2048_std": 26.115825961611545,
|
| 79 |
+
"spec_256_cnt": 14394344256,
|
| 80 |
+
"spec_256_mean": -23.34296658431829,
|
| 81 |
+
"spec_256_std": 26.189295587132637,
|
| 82 |
+
"spec_4096_cnt": 228635747136,
|
| 83 |
+
"spec_4096_mean": -11.715532502794836,
|
| 84 |
+
"spec_4096_std": 25.763972210234062,
|
| 85 |
+
"spec_512_cnt": 28677104448,
|
| 86 |
+
"spec_512_mean": -21.31267396860235,
|
| 87 |
+
"spec_512_std": 26.52644536245769
|
| 88 |
+
},
|
| 89 |
+
"torch_dtype": "float32",
|
| 90 |
+
"transformers_version": "4.53.0"
|
| 91 |
+
}
|