知识卡片

普通读书笔记卡 · 6700.d · 生成模型:GPT-6

加载器不能消除预先物化的内存

理解等级:中级

把全部窗口对先展开成Python列表,会在创建Dataset时占据与训练对数量相当的内存。DataLoader仅分批读取已存列表,不能让先前全量list或词表构造自动流式;大语料需从生成与采样阶段就设计迭代访问。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      ... self.data = [i for s in dataset['moving_window'] for i in s] ... ... def __len__(self): ... return len(self.data) ... ... def __getitem__(self, idx): ... return self.data[idx] And your DataLoader will take care of memory management for you. This will ensure your pipeline is reusable for virtually any size corpus, even all of Wikipedia: from torch.utils.data import DataLoader dataloader = {} for k in dset.keys():

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第06章《用词嵌入推理》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-6.html) - 结论依据:摘录中关于加载器不能消除预先物化的内存的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:... self.data = [i for s in dataset['moving_window'] for i in s] ... ... def __len__(self): ... return len(self.data) ... ... def __getitem__(self, idx): ... return self.data[idx] And your DataLoader will take care of memory management for you. This will ensure your pipeline is reusable for virtually any size corpus, even all of Wikipedia: from torch.utils.data import DataLoader dataloader = {} for k in dset.keys():