pytorch/text

Vocab vectors using complete pretrained-embedding?

Open

#446 创建于 2018年10月12日

在 GitHub 查看
 (6 评论) (0 反应) (0 负责人)Python (3,396 star) (822 fork)batch import
enhancementhelp wanted

描述

I am new to pytorch and nlp. I have a question when I tried to build a model.

Since my training dataset is not so big, the size of its vocab is relatively small (around 5000). However, I want to deal with any other user input which could be out of this vocabulary.

The problem is, in the model I trained, the embedding layer's weight is based on the vectors of the field, not the whole word2vec pretrained embeddings. So I cannot modified it after the training is done.

I wondered is there any better approach to do it? Thanks in advance!

贡献者指南

技术栈
python
领域
machine learning
议题类型
feature
难度面向新贡献者的预计实现难度，1 表示很小改动，5 表示专家级工作。
3
预计时间有经验贡献者完成调查、实现、测试并准备 pull request 的粗略时间范围。
half day
活动状态议题当前的可参与程度：新鲜、活跃、陈旧、阻塞或等待维护者输入。
stale
清晰度议题是否清楚说明期望改动、验收标准和下一步。
clear
前置要求
PyTorch basicsword embeddings
新手友好度1-100 的估计分数，表示该议题对首次贡献者的友好程度。
50
研究方向
Investigate how PyTorch's torchtext handles vocabulary and embedding layers. Look into existing discussion on this issue (e.g., comments on GitHub issue #446) and evaluate if a utility function for loading full pretrained embeddings is feasible. Consider adding a method to allow users to specify a pretrained embedding source that maps tokens not in the training vocab to their vectors from a larger set.