GexLoong

> I've received your dataset, it very helpful, thanks a lot. But I found that some training data slices don't have masks, like 'case0005_slice000.npz' to 'case0005_slice022.npz', I think these slices...

> my machine got the error about " out of memory"  > > i have change the "total_batch_size" to 2， ues "amp-fp16" ， and use dataset "'2018-01-01' to '2018-01-10'"...

我的环境配置： * python 3.7.16 * torch 1.10.0 * torchvision 0.10.0 * mmcv-full 1.6.1 torch1.10

> @MyGitHub-G 我看图片里的报错是没有指明work-dir这个参数 这个应该不是没有指明work-dir参数的问题，代码里面如果没有指明会有默认目录创建。这个问题貌似是程序运行的问题，我如果不用sh运行，直接运行train.py，会报错core dump