deep-learning - problem on training(first epoch take a long time more than 4 hours )using colab and fastai
问题描述
I have a problem on training (first epoch takes a long time more than 4 hours) using colab and fastai. It takes me about 5 hours only for the first epoch which is not useful using google colab because the limitation using of GPU I'm using 'efficientnet-b4' and about 52k training photos after data augmentation and 11k valid photos.
The problem is shown on the photo:
解决方案
Colab takes alot of time fetching the data from google drive as it is not indexed efficiently and thus makes the first epoch very long. For large datasets, I would recommend using google cloud storage and importing data from there.
推荐阅读
- python - 多图下载
- python - 如何从写入数组的文本文件中获取数组?
- javascript - ReactJS:将函数传递给子组件会导致 TypeError
- python - 获取 Shapely 多边形内具有整数坐标的所有点
- python - 使用 pdfplumber 和 Paramiko 从 SFTP 服务器读取 PDF 文件
- flutter - 如何在颤动中等待常规(非异步)代码(函数)?
- python - PYTHON -- ModuleNotFoundError: 当我尝试导入包时没有名为“playsound”的模块
- javascript - 开发 JQuery Query Builder 接口的克隆
- c# - 对象数组未绑定的 Asp.net Core 3.1 Web api 请求
- excel - POI:使用具有条件格式的 Excel 模板