pytorch/ignite

Add no improvement handler (similar to early stopping handler)

Open

#2,314 建立於 2021年11月9日

在 GitHub 查看
 (8 留言) (0 反應) (1 負責人)Python (4,313 star) (602 fork)batch import
enhancementhelp wanted

描述

🚀 Feature

To add a 'no-improvement' handler. It's similar to early stopping, except that early stopping calls the training to terminate, 'no-improvement' handler would've calls a function that user pass in. User could pass in different functions that they want to execute if they find no improvements is made. They may want to try various things before terminate the training.

貢獻者指南

技術棧
pythonpytorch
領域
machine learning
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。
blocked
清晰度議題是否清楚說明預期改動、驗收標準和下一步。
clear
前置要求
Knowledge of PyTorch Ignite handlersUnderstanding of early stopping handler
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
Examine the existing early stopping handler implementation in PyTorch Ignite (likely in ignite/handlers/early stopping.py). Understand the callback interface and training termination flow. Review comments in issue #2314 for any user proposed function signatures. The goal is to create a handler that triggers a user defined function when no improvement is observed, rather than terminating training. Implementation should mirror early stopping's logic but invoke the callback instead of stopping.
Add no improvement handler (similar to early stopping handler) · pytorch/ignite#2314 | Good First Issue