import torch from collections import defaultdict from torch.utils.data import Dataset, DataLoader, Sampler class SequenceDataset(Dataset): def __init__(self, config, sequences, indexes): self.sequences = sequences self.indexes = indexes self.config = config self.index_to_samples = defaultdict(list) for i, index in enumerate(indexes): self.index_to_samples[index].append(i) def __len__(self): return len(self.sequences) def __getitem__(self, idx): seq = self.sequences[idx] index = self.indexes[idx] item_seq = seq[:-1] labels = seq[-1] seq_length = len(item_seq) padding_length = self.config['max_seq_length'] - len(item_seq) if padding_length > 0: item_seq = item_seq + [0] * padding_length # 在后面填充0 return { 'item_seqs': torch.tensor(item_seq, dtype=torch.long), 'labels': torch.tensor(labels, dtype=torch.long), 'seq_lengths': seq_length, 'idx': index } class NormalRecData: def __init__(self, config: dict): self.config = config def load_data(self): from pathlib import Path source_dict = { 'C': 'CDs_and_Vinyl', 'O': 'Sports_and_Outdoors', 'T': 'Toys_and_Games', } self.config['source_dict'] = source_dict def read_data_from_file(domain, mode=''): # 基础路径和域名映射字典 base_path = Path('data/') # 获取目标文件路径 file_path = base_path / source_dict[domain] / '{}data.txt'.format(mode) with file_path.open('r') as file: item_seqs = [list(map(int, line.split())) for line in file] if mode == '': flat_list = [item for sublist in item_seqs for item in sublist] import numpy as np item_num = np.max(flat_list) return item_seqs, item_num else: return item_seqs total_item_num = 0 all_data = [] train_data = [] valid_data = [] test_data = [] train_index = [] valid_index = [] test_index = [] select_pool = None for index, key in enumerate(source_dict.keys()): if self.config.get('ab', None) == 'single': if key != self.config['sd']: continue tmp_item_seqs, cur_item_num = read_data_from_file(key) all_data.extend([[item + total_item_num for item in sublist] for sublist in tmp_item_seqs]) tmp_train_item_seqs, tmp_valid_item_seqs, tmp_test_item_seqs = (read_data_from_file(key, mode='train_'), read_data_from_file(key, mode='valid_'), read_data_from_file(key, mode='test_')) tmp_train_item_seqs = [[item + total_item_num for item in sublist] for sublist in tmp_train_item_seqs] tmp_valid_item_seqs = [[item + total_item_num for item in sublist] for sublist in tmp_valid_item_seqs] tmp_test_item_seqs = [[item + total_item_num for item in sublist] for sublist in tmp_test_item_seqs] tmp_train_idx = [index] * len(tmp_train_item_seqs) tmp_valid_idx = [index] * len(tmp_valid_item_seqs) tmp_test_idx = [index] * len(tmp_test_item_seqs) if key == self.config['td']: valid_data.extend(tmp_valid_item_seqs) valid_index.extend(tmp_valid_idx) test_data.extend(tmp_test_item_seqs) test_index.extend(tmp_test_idx) select_pool = [total_item_num + 1, total_item_num + cur_item_num + 1] if key in self.config['sd']: train_data.extend(tmp_train_item_seqs) train_index.extend(tmp_train_idx) total_item_num += cur_item_num return (SequenceDataset(self.config, train_data, train_index), SequenceDataset(self.config, valid_data, valid_index), SequenceDataset(self.config, test_data, test_index), select_pool, total_item_num) # train_data[111] # test_data[130]