|
| 1 | +import gin |
| 2 | +import json |
| 3 | +import random |
| 4 | +import traceback |
| 5 | +from pathlib import Path |
| 6 | +from typing import Union |
| 7 | + |
| 8 | +import torch |
| 9 | +import pytorch_lightning as L |
| 10 | +import yaml |
| 11 | +from tqdm import tqdm |
| 12 | +from torch.utils.data import DataLoader |
| 13 | + |
| 14 | + |
| 15 | +from .data_utils import AudioDataset |
| 16 | + |
| 17 | + |
| 18 | +@gin.configurable |
| 19 | +class DiscotubeTextAudioDataset(AudioDataset): |
| 20 | + """Generic audio dataset.""" |
| 21 | + |
| 22 | + def __init__( |
| 23 | + self, |
| 24 | + data_dir: Path, |
| 25 | + filelist: Path, |
| 26 | + metadata_youtube: dict, |
| 27 | + metadata_discogs: dict, |
| 28 | + metadata_id_map: dict, |
| 29 | + frame_offset: Union[int, str] = "random", |
| 30 | + ): |
| 31 | + super().__init__( |
| 32 | + data_dir=data_dir, |
| 33 | + filelist=filelist, |
| 34 | + frame_offset=frame_offset, |
| 35 | + ) |
| 36 | + |
| 37 | + self.metadata_youtube = metadata_youtube |
| 38 | + self.metadata_discogs = metadata_discogs |
| 39 | + self.metadata_id_map = metadata_id_map |
| 40 | + |
| 41 | + def __len__(self): |
| 42 | + return len(self.filelist) |
| 43 | + |
| 44 | + @staticmethod |
| 45 | + def get_audio_path(youtube_id: str) -> Path: |
| 46 | + return Path(youtube_id[:2], youtube_id).with_suffix(".mmap") |
| 47 | + |
| 48 | + def __getitem__(self, idx): |
| 49 | + try: |
| 50 | + id_yt = self.filelist[idx] |
| 51 | + |
| 52 | + file_path = self.data_dir / self.get_audio_path(id_yt) |
| 53 | + |
| 54 | + # load audio |
| 55 | + # audio = self.load_audio(file_path, frame_offset=self.frame_offset) |
| 56 | + audio = torch.rand(1, 16000 * 30) |
| 57 | + |
| 58 | + # load YouTube metadata |
| 59 | + meta_youtube = self.metadata_youtube[id_yt] |
| 60 | + |
| 61 | + # load discogs metadata |
| 62 | + ids_discogs = self.metadata_id_map[id_yt] |
| 63 | + |
| 64 | + # sample randonly among available releases |
| 65 | + id_discogs = random.choice(ids_discogs) |
| 66 | + meta_discogs = self.metadata_discogs[id_discogs] |
| 67 | + |
| 68 | + # process metadata |
| 69 | + text = self.preprocess_text( |
| 70 | + {"youtube_metadata": meta_youtube, "discogs_metadata": meta_discogs} |
| 71 | + ) |
| 72 | + except Exception: |
| 73 | + print(f"Error loading {self.filelist[idx]}") |
| 74 | + print(traceback.format_exc()) |
| 75 | + return [None, None] |
| 76 | + |
| 77 | + return [audio, text] |
| 78 | + |
| 79 | + def preprocess_text(self, metadata: dict) -> str: |
| 80 | + """Text preprocessing""" |
| 81 | + |
| 82 | + # Process YouTube metadata |
| 83 | + fields_to_keep = ["description", "categories", "tags", "view_count"] |
| 84 | + youtube_metadata = metadata["youtube_metadata"] |
| 85 | + new_youtube_metadata = { |
| 86 | + field: youtube_metadata[field] |
| 87 | + for field in fields_to_keep |
| 88 | + if field in youtube_metadata |
| 89 | + } |
| 90 | + |
| 91 | + # Process Discogs metadata |
| 92 | + |
| 93 | + fields_to_keep = ["labels", "genres", "styles", "country", "released"] |
| 94 | + dicogs_metadata = metadata["discogs_metadata"] |
| 95 | + new_discogs_metadata = { |
| 96 | + field: dicogs_metadata[field] |
| 97 | + for field in fields_to_keep |
| 98 | + if field in dicogs_metadata |
| 99 | + } |
| 100 | + |
| 101 | + # Fetch artist description |
| 102 | + # TODO: Get this too |
| 103 | + |
| 104 | + metadata = { |
| 105 | + "youtube_metadata": new_youtube_metadata, |
| 106 | + "discogs_metadata": new_discogs_metadata, |
| 107 | + } |
| 108 | + |
| 109 | + # format as YAML |
| 110 | + yaml_text = yaml.dump(metadata, sort_keys=False) |
| 111 | + return yaml_text |
| 112 | + |
| 113 | + |
| 114 | +@gin.configurable |
| 115 | +class DiscotubeTextAudioDataModule(L.LightningDataModule): |
| 116 | + """AudioDataModule for the Discogs dataset.""" |
| 117 | + |
| 118 | + def __init__( |
| 119 | + self, |
| 120 | + batch_size: int, |
| 121 | + data_dir: Path, |
| 122 | + filelist_train: Path, |
| 123 | + filelist_val: Path, |
| 124 | + metadata_youtube_file: Path, |
| 125 | + metadata_discogs_file: Path, |
| 126 | + metadata_id_map_file: Path, |
| 127 | + num_workers: int, |
| 128 | + ): |
| 129 | + super().__init__() |
| 130 | + |
| 131 | + self.batch_size = batch_size |
| 132 | + |
| 133 | + self.data_dir = Path(data_dir) |
| 134 | + self.filelist_train = Path(filelist_train) |
| 135 | + self.filelist_val = Path(filelist_val) |
| 136 | + |
| 137 | + self.num_workers = num_workers |
| 138 | + |
| 139 | + self.metadata_youtube_file = metadata_youtube_file |
| 140 | + self.metadata_discogs_file = metadata_discogs_file |
| 141 | + self.metadata_id_map_file = metadata_id_map_file |
| 142 | + |
| 143 | + def setup(self, stage: str): |
| 144 | + # load YouTube metadata from jsonl (one json object per line) |
| 145 | + self.metadata_youtube = dict() |
| 146 | + with open(self.metadata_youtube_file, "r") as f: |
| 147 | + for line in tqdm(f.readlines(), desc="Loading YouTube metadata"): |
| 148 | + line = json.loads(line) |
| 149 | + self.metadata_youtube[line["id"]] = line |
| 150 | + |
| 151 | + # load Discogs metadata from jsonl (one json object per line) |
| 152 | + self.metadata_discogs = dict() |
| 153 | + with open(self.metadata_discogs_file, "r") as f: |
| 154 | + for line in tqdm(f.readlines(), desc="Loading Discogs metadata"): |
| 155 | + line = json.loads(line) |
| 156 | + self.metadata_discogs[line["id"]] = line |
| 157 | + |
| 158 | + # load the id map from jsonl (one json object per line) |
| 159 | + self.metadata_id_map = dict() |
| 160 | + with open(self.metadata_id_map_file, "r") as f: |
| 161 | + for line in tqdm(f.readlines(), desc="Loading ID map"): |
| 162 | + line = json.loads(line) |
| 163 | + for k, v in line.items(): |
| 164 | + self.metadata_id_map[k] = v |
| 165 | + |
| 166 | + self.dataset_train = DiscotubeTextAudioDataset( |
| 167 | + self.data_dir, |
| 168 | + filelist=self.filelist_train, |
| 169 | + metadata_youtube=self.metadata_youtube, |
| 170 | + metadata_discogs=self.metadata_discogs, |
| 171 | + metadata_id_map=self.metadata_id_map, |
| 172 | + ) |
| 173 | + self.dataset_val = DiscotubeTextAudioDataset( |
| 174 | + self.data_dir, |
| 175 | + filelist=self.filelist_val, |
| 176 | + metadata_youtube=self.metadata_youtube, |
| 177 | + metadata_discogs=self.metadata_discogs, |
| 178 | + metadata_id_map=self.metadata_id_map, |
| 179 | + ) |
| 180 | + |
| 181 | + def train_dataloader(self): |
| 182 | + return DataLoader( |
| 183 | + self.dataset_train, |
| 184 | + batch_size=self.batch_size, |
| 185 | + num_workers=self.num_workers, |
| 186 | + pin_memory=True, |
| 187 | + ) |
| 188 | + |
| 189 | + def val_dataloader(self): |
| 190 | + return DataLoader( |
| 191 | + self.dataset_val, |
| 192 | + batch_size=self.batch_size, |
| 193 | + num_workers=self.num_workers, |
| 194 | + pin_memory=True, |
| 195 | + ) |
0 commit comments