-
As you know, inference time depends on the hardware performance. For example, the fps of retinanet fp16 model is 31.6 fps, but it is not easy to achieve a similar speed. |
Beta Was this translation helpful? Give feedback.
Replies: 1 comment 1 reply
-
I guess you could find the hardware info in the training log, which was listed in the rightest column of a performance table. See here for instance: https://download.openmmlab.com/mmdetection/v2.0/fp16/retinanet_r50_fpn_fp16_1x_coco/retinanet_r50_fpn_fp16_1x_coco_20200702_020127.log.json |
Beta Was this translation helpful? Give feedback.
I guess you could find the hardware info in the training log, which was listed in the rightest column of a performance table. See here for instance: https://download.openmmlab.com/mmdetection/v2.0/fp16/retinanet_r50_fpn_fp16_1x_coco/retinanet_r50_fpn_fp16_1x_coco_20200702_020127.log.json