FluxML/Flux.jl

Overloading Flux.batch for custom eltypes

Open

#1,821 创建于 2022年1月3日

在 GitHub 查看
 (13 评论) (0 反应) (0 负责人)Julia (4,725 star) (619 fork)batch import
enhancementgood first issuehelp wanted

描述

I ran into a use case for the batch function to operate on generators. I noticed that the batch function is not all that compatible with generators since it doesn't utilize the collect function for them. I'm curious if this additional functionality for the batch function is something that Flux.jl is interested in?

贡献者指南

技术栈
无
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。
1-2 days
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。
needs maintainer response
清晰度议题是否清楚说明期望改动、验收标准和下一步。
mostly clear
前置要求
Familiarity with JuliaUnderstanding of Flux's batch function
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。
25
研究方向
Investigate the current implementation of `Flux.batch` in the Flux.jl repository, likely in files like `src/data/batch.jl`. Understand how it handles different element types and why it does not work with generators. Consider how to overload or extend `batch` to support custom eltypes, possibly by using `collect` on generators. Review the comments in the issue for any suggested approaches or maintainer input.