gib专题

RuntimeError: CUDA out of memory. Tried to allocate 1.77 GiB?如何解决

🏆本文收录于「Bug调优」专栏,主要记录项目实战过程中的Bug之前因后果及提供真实有效的解决方案,希望能够助你一臂之力,帮你早日登顶实现财富自由🚀;同时,欢迎大家关注&&收藏&&订阅!持续更新中,up!up!up!! 问题描述   mmdetection运行mask rcnn,训练模型时运行train.py出现RuntimeError: CUDA out of memory. T

GB和GiB的区别

KB 是 kilobyte 的缩写; MB 是 megabyte 的缩写; GB 是 gigabyte 的缩写; TB 是 terabyte 的缩写; PB 是 petabyte 的缩写 这些是十进制的容量单位。 其中: 1 PB = 1,000 TB1 TB = 1,000 GB1 GB = 1,000 MB1 MB = 1,000 KB1 KB = 1,000 Byte KiB

GiB is reserved by PyTorch but unallocated

碰到错误 Of the allocated memory 12.53 GiB is allocated by PyTorch, and 1.22 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid