AlibabaResearch / AlibabaResearch/AdvancedLiterateMachinery

VGT training memory requirement

未关闭
#198 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
C++
星标
1.8k
派生
195
PR 合并指标
30 天内没有已合并 PR

描述

I am able to train it with one 2080Ti with 12 GB of VRAM but I would have to resort to having a batch size of 1.

Training with a batch size of 4 and 4 2080Tis throws out of memory after some time even with mixed precision enabled.

Please let me know about the:
1) Normal memory requirements for training
2) Any ways to reduce the memory required.

Thanks very much in advance.

贡献指南

这个仓库没有索引到贡献指南

评估

这个 Issue 还没有评估数据。

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。