pytorch/ignite

Debug mode for ignite

Open

#1,992 建立於 2021年5月7日

在 GitHub 查看
 (15 留言) (2 反應) (0 負責人)Python (4,313 star) (602 fork)batch import
enhancementhelp wanted

描述

🚀 Feature

The idea is to provide a debug mode as discussed here: https://github.com/pytorch/ignite/issues/1989#issuecomment-833643158.

All the possible options can be discussed here. We could add an helper to track the workflow of handlers from the engine run. On the other hand, some specific debug tools are needed to track the values of PyTorch objects (LR, loss, grads, etc.). Note that the preferred way to extend ignite is handlers' design so let's create a new set of tools for the debugging.

貢獻者指南

技術棧
python
領域
machine learningdeveloper experience
議題類型
feature
難度面向新貢獻者的預計實作難度，1 表示很小改動，5 表示專家級工作。
3
預計時間有經驗貢獻者完成調查、實作、測試並準備 pull request 的粗略時間範圍。
1-2 days
活動狀態議題目前的可參與程度：新鮮、活躍、陳舊、阻塞或等待維護者輸入。
stale
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
mostly clear
前置要求
PythonPyTorchIgniteDebugging
新手友善度1-100 的估計分數，表示該議題對首次貢獻者的友善程度。
50
研究方向
Review the referenced discussion in issue #1989 for context on the desired debug features. Analyze the Ignite handler pattern to design a debug handler that logs PyTorch objects like loss and gradients. Consider creating a separate module for debugging tools that can be attached to the engine. The implementation should follow the existing handler design to maintain consistency.