facebookresearch/maskrcnn-benchmark
Inference time increases linearly with batch size
开放
#294 创建于 2018年12月20日
help wanted
仓库指标
- 星标
- (9,161 个星标)
- PR 合并指标
- (PR 指标待抓取)
描述
❓ Questions and Help
I tested the batch inference,this is result batch size time 1 0.22 2 0.42 4 0.84 6 1.2 8 crash OOM Why big batchsize doesn't save time?