知识卡片

普通读书笔记卡 · 6746.d · 生成模型:GPT-6

文本编码接口要核对返回对象

理解等级:中级

ByteLevelBPETokenizer的编码对象、ID列表与模型整型张量承担不同角色。预处理应明确取ids、打包源目标和padding,不能把tokenizer对象当任意可调用函数;source_lang等变量也须与实际SRC配置统一。

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

      You can use your English tokenizer to build a preprocessing function that flattens the Dataset structure and returns a list of lists of token IDs (without padding): def preprocess(examples): src = [x[source_lang] for x in examples["translation"]] src_toks = [tokenize_src(x) for x in src] # tgt = [x[target_lang] for x in examples["translation"]] # tgt_toks = [tokenize_tgt(x) for x in tgt] return src_toks The TranslationTransformer model At

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第09章《可堆叠的Transformer》(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-9.html) - 结论依据:摘录中关于文本编码接口要核对返回对象的定义、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:You can use your English tokenizer to build a preprocessing function that flattens the Dataset structure and returns a list of lists of token IDs (without padding): def preprocess(examples): src = [x[source_lang] for x in examples["translation"]] src_toks = [tokenize_src(x) for x in src] # tgt = [x[target_lang] for x in examples["translation"]] # tgt_toks = [tokenize_tgt(x) for x in tgt] return src_toks The TranslationTransformer model At