SemantiCodec: An Ultra Low Bitrate Semantic Audio Codec for General Sound
Paper • 2405.00233 • Published • 17
audio_tokens sequencelengths 64 3.01k | genre_id int64 0 18 | genre stringclasses 19
values | song_id int64 0 25k |
|---|---|---|---|
[
569,
1834,
2459,
7787,
2947,
3779,
288,
4151,
288,
4245,
288,
493,
708,
3028,
873,
3168,
902,
1295,
3270,
6808,
1132,
2185,
3344,
1745,
3301,
7344,
490,
5469,
1080,
5853,
40,
636,
711,
5759,
970,
5761,
1132,
2378,
3820,
7700,
693,
586... | 0 | Electronic | 0 |
[
1633,
6381,
2793,
6385,
1035,
429,
3132,
1388,
3222,
7374,
1331,
1455,
3007,
6419,
3283,
7895,
2088,
937,
3562,
1917,
1783,
3202,
3991,
1958,
1167,
8175,
1331,
4389,
1981,
5864,
1114,
6751,
1673,
4389,
3322,
7025,
2013,
6959,
3132,
3487,
... | 9 | International | 1 |
[
2636,
7165,
2223,
6240,
2175,
721,
1853,
2300,
2962,
66,
3316,
3447,
1740,
2071,
1770,
7988,
2088,
859,
2626,
1790,
3808,
471,
3005,
7602,
3364,
7410,
1311,
1712,
2836,
3846,
3648,
4324,
1673,
2245,
3214,
6410,
3029,
237,
3635,
6722,
15... | 6 | Chiptune / Glitch | 2 |
[
1287,
7427,
2676,
1153,
2993,
4972,
2255,
4483,
2318,
3963,
1195,
7716,
1505,
3525,
1999,
6531,
1302,
2908,
2000,
2635,
3546,
5038,
1447,
7011,
1968,
1966,
1934,
464,
3409,
2810,
3138,
1846,
3560,
7092,
2000,
897,
2893,
878,
1447,
1623,
... | 4 | Hip-Hop | 4 |
[
2012,
838,
2012,
4268,
1470,
4890,
1809,
587,
2020,
7961,
1647,
5088,
1786,
7284,
2242,
4325,
1692,
1360,
1434,
5276,
1355,
6022,
1204,
2081,
1204,
207,
1575,
4261,
1921,
3531,
3474,
5978,
1692,
5348,
1525,
1339,
1720,
809,
1541,
7015,
... | 14 | Country | 5 |
[
2783,
4287,
2826,
1172,
2702,
7043,
2702,
6287,
2826,
5083,
2826,
1620,
2826,
3063,
2826,
5392,
2783,
4171,
3677,
2393,
3747,
1343,
2518,
4457,
3460,
6689,
3647,
1021,
2641,
1960,
2599,
8076,
2783,
7497,
2497,
6531,
1844,
6004,
3107,
7726... | 12 | Old-Time / Historic | 7 |
[
1691,
7207,
880,
6933,
825,
4741,
1470,
5301,
606,
3377,
286,
5675,
141,
3316,
517,
547,
1184,
3502,
145,
1689,
3883,
4778,
208,
7400,
606,
4881,
1646,
4986,
1934,
1153,
517,
6135,
3162,
6767,
1932,
6260,
3883,
6719,
474,
7042,
3798,
... | 4 | Hip-Hop | 9 |
[
1416,
4748,
2016,
4546,
1074,
1706,
2399,
7558,
1532,
6109,
1381,
3634,
3770,
8182,
3728,
5238,
1424,
3698,
3293,
2584,
3319,
3213,
2613,
6737,
3357,
2798,
2418,
6842,
3770,
3405,
1237,
2625,
1657,
2735,
2603,
3240,
3319,
1878,
2613,
2040... | 2 | Punk | 10 |
[
1405,
1593,
2716,
515,
4056,
5544,
3922,
4070,
1034,
4650,
2557,
562,
2321,
574,
3038,
2196,
2556,
5982,
2716,
5263,
3619,
7228,
2769,
205,
3480,
6752,
2489,
1027,
2171,
2330,
3485,
5177,
3917,
3171,
2716,
5057,
2464,
7119,
3096,
621,
3... | 0 | Electronic | 12 |
[
1880,
3951,
2310,
7174,
3686,
520,
2133,
3057,
2555,
6257,
2090,
4877,
3574,
7387,
3182,
7481,
1880,
4430,
3362,
6982,
1476,
2430,
1476,
2255,
2511,
3345,
723,
5986,
2224,
1747,
598,
3646,
2331,
1556,
3767,
2069,
2077,
1626,
2399,
8100,
... | 7 | Instrumental | 13 |
[
2728,
3967,
2279,
5815,
3860,
4260,
3337,
7949,
2921,
7474,
2407,
8068,
2192,
3061,
3188,
6434,
2728,
883,
3341,
654,
3860,
5029,
3337,
4386,
2921,
461,
2407,
4951,
2140,
3294,
3188,
4390,
2728,
7594,
3341,
3195,
1694,
7421,
3337,
8138,
... | 3 | Experimental | 14 |
[
2669,
2463,
2826,
1259,
2119,
7063,
2669,
5950,
984,
1674,
2669,
6908,
2826,
6833,
2826,
1101,
2704,
86,
2550,
983,
2119,
3883,
2932,
1321,
2673,
5369,
3089,
791,
3339,
2256,
2109,
205,
4061,
5245,
3744,
30,
2119,
7145,
2435,
2107,
3405... | 17 | Blues | 15 |
[
3555,
6205,
3203,
7594,
1133,
7005,
1175,
5983,
1679,
5373,
1679,
4993,
2485,
3942,
1060,
7794,
1493,
7956,
1629,
2298,
1133,
3547,
1175,
2661,
1557,
7532,
1679,
4219,
2018,
3054,
1576,
1223,
3555,
8146,
1253,
6610,
1133,
2425,
1175,
7702... | 4 | Hip-Hop | 16 |
End of preview. Expand in Data Studio
This dataset contains tokenized audio from lewtun/music_genres using SemantiCodec for performing experiments on AR music generation
The following script is used for tokenization
from datasets import load_dataset, Dataset, DatasetDict
model_id = ""
repo_name = ""
user_name = ""
token = ""
cache_dir = "cache"
vocab_size = 4096
dataset = load_dataset(model_id, cache_dir=cache_dir, trust_remote_code=True)
from semanticodec import SemantiCodec
semanticodec = SemantiCodec(token_rate=100, semantic_vocab_size=vocab_size)
import soundfile as sf
from tqdm import tqdm
import math
dd = {
"train": 0,
"test": 0
}
for split in ["train", "test"]:
tkns = []
for idx in tqdm(range(len(dataset[split]))):
sample = dataset[split][idx]["audio"]
array = sample["array"]
sr = sample["sampling_rate"]
sf.write("output.wav", array, sr)
tokens = semanticodec.encode("output.wav").detach().cpu().numpy().flatten()
tkns.append(tokens)
ds = Dataset.from_dict({
"audio_tokens": tkns,
"genre_id": list(dataset[split]["genre_id"]),
"genre": list(dataset[split]["genre"]),
"song_id": list(dataset[split]["song_id"])
})
dd[split] = ds
dd = DatasetDict(dd)
dd.save_to_disk(repo_name)
dd.push_to_hub(f"{user_name}/{repo_name}", token=token)
- Downloads last month
- 4