pytorch/examples

Resume will override learning rate on command line.

Open

#816 创建于 2020年8月19日

在 GitHub 查看
 (1 评论) (0 反应) (0 负责人)Python (21,634 star) (9,429 fork)batch import
good first issue

描述

If you resume from a checkpoint and want to run with a new learning rate, the saved learning rate will overwrite what you specify on the command line. The fix is simple.

贡献者指南

技术栈
pythonpytorch
领域
machine learning
议题类型
bug
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。
2
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。
under 1 hour
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
Python programmingPyTorch basics
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。
70
研究方向
Look at the training script where checkpoint loading is implemented. The issue is that the saved learning rate from the checkpoint overwrites the command line argument. Identify the line where the learning rate is restored and modify it to only set the learning rate from the checkpoint if no command line argument is provided. Ensure that the command line value takes precedence.