V100无法复现您的帧率

[jiajia131]

Originally posted by @junjiehe96 in #37 (comment)

        +---------------------------------------------------------------------------------------+

| NVIDIA-SMI 530.30.02 Driver Version: 530.30.02 CUDA Version: 12.1 |
|-----------------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+======================+======================|
| 0 Tesla V100-PCIE-32GB On | 00000000:07:00.0 Off | 0 |
| N/A 31C P0 37W / 250W| 5146MiB / 32768MiB | 4% Default |
| | | N/A |
+-----------------------------------------+----------------------+----------------------+

[04/06 17:46:17 d2.evaluation.evaluator]: Inference done 370/5000. Dataloading: 0.0022 s/iter. Inference: 0.0564 s/iter. Eval: 0.1694 s/iter. Total: 0.2281 s/iter. ETA=0:17:36
[04/06 17:46:22 d2.evaluation.evaluator]: Inference done 391/5000. Dataloading: 0.0022 s/iter. Inference: 0.0567 s/iter. Eval: 0.1699 s/iter. Total: 0.2289 s/iter. ETA=0:17:35
[04/06 17:46:27 d2.evaluation.evaluator]: Inference done 412/5000. Dataloading: 0.0022 s/iter. Inference: 0.0568 s/iter. Eval: 0.1704 s/iter. Total: 0.2296 s/iter. ETA=0:17:33
[04/06 17:46:32 d2.evaluation.evaluator]: Inference done 433/5000. Dataloading: 0.0023 s/iter. Inference: 0.0567 s/iter. Eval: 0.1712 s/iter. Total: 0.2303 s/iter. ETA=0:17:31
[04/06 17:46:37 d2.evaluation.evaluator]: Inference done 457/5000. Dataloading: 0.0022 s/iter. Inference: 0.0565 s/iter. Eval: 0.1706 s/iter. Total: 0.2295 s/iter. ETA=0:17:22
[04/06 17:46:42 d2.evaluation.evaluator]: Inference done 480/5000. Dataloading: 0.0022 s/iter. Inference: 0.0564 s/iter. Eval: 0.1704 s/iter. Total: 0.2292 s/iter. ETA=0:17:16

        请问，为什么我用V100无法复现你的帧率？完全使用fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml默认设置只有不到20帧。

[jiajia131]

[04/07 04:39:41 d2.evaluation.evaluator]: Total inference time: 0:16:15.725855 (0.195341 s / iter per device, on 1 devices)
[04/07 04:39:41 d2.evaluation.evaluator]: Total inference pure compute time: 0:04:38 (0.055843 s / iter per device, on 1 devices)

[junjiehe96]

能提供你的详细测试脚本和log日志吗

[jiajia131]

能提供你的详细测试脚本和log日志吗

log.txt
python train_net.py --eval-only --num-gpus 1 --config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml MODEL.WEIGHTS ../fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth

谢谢您，这是使用 Tesla V100-SXM2-16GB

[junjiehe96]

能提供你的详细测试脚本和log日志吗

log.txt python train_net.py --eval-only --num-gpus 1 --config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml MODEL.WEIGHTS ../fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth

谢谢您，这是使用 Tesla V100-SXM2-16GB

您能用下面这条命令再测一下吗
CUDA_VISIBLE_DEVICES=0 python tools/analyze_model.py --tasks speed
--config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml

[jiajia131]

能提供你的详细测试脚本和log日志吗

log.txt python train_net.py --eval-only --num-gpus 1 --config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml MODEL.WEIGHTS ../fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth
谢谢您，这是使用 Tesla V100-SXM2-16GB

您能用下面这条命令再测一下吗 CUDA_VISIBLE_DEVICES=0 python tools/analyze_model.py --tasks speed --config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml

CUDA_VISIBLE_DEVICES=0 python tools/analyze_model.py --tasks speed
--config-file ./configs/coco/instance-segmentation/fastinst_R50-vd-dcn_ppm-fpn_x3_640.yaml
MODEL.WEIGHTS ./fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth

用这个会更慢了
[04/07 06:20:28 d2.checkpoint.detection_checkpoint]: [DetectionCheckpointer] Loading from /hy-tmp/FastInst-main-240320/fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth ...
[04/07 06:20:28 fvcore.common.checkpoint]: [Checkpointer] Loading from /hy-tmp/FastInst-main-240320/fastinst_R50-vd-dcn_ppm-fpn_x3_640_40.1.pth ...
[04/07 06:20:28 d2.evaluation.evaluator]: Start inference on 5000 batches
[04/07 06:20:31 d2.evaluation.evaluator]: Inference done 11/5000. Dataloading: 0.0149 s/iter. Inference: 0.1241 s/iter. Eval: 0.0000 s/iter. Total: 0.1391 s/iter. ETA=0:11:33
[04/07 06:20:36 d2.evaluation.evaluator]: Inference done 47/5000. Dataloading: 0.0247 s/iter. Inference: 0.1146 s/iter. Eval: 0.0000 s/iter. Total: 0.1394 s/iter. ETA=0:11:30
[04/07 06:20:41 d2.evaluation.evaluator]: Inference done 84/5000. Dataloading: 0.0277 s/iter. Inference: 0.1118 s/iter. Eval: 0.0000 s/iter. Total: 0.1396 s/iter. ETA=0:11:26
[04/07 06:20:46 d2.evaluation.evaluator]: Inference done 120/5000. Dataloading: 0.0276 s/iter. Inference: 0.1127 s/iter. Eval: 0.0000 s/iter. Total: 0.1404 s/iter. ETA=0:11:24
[04/07 06:20:51 d2.evaluation.evaluator]: Inference done 155/5000. Dataloading: 0.0293 s/iter. Inference: 0.1122 s/iter. Eval: 0.0000 s/iter. Total: 0.1416 s/iter. ETA=0:11:25
[04/07 06:20:56 d2.evaluation.evaluator]: Inference done 191/5000. Dataloading: 0.0298 s/iter. Inference: 0.1117 s/iter. Eval: 0.0000 s/iter. Total: 0.1416 s/iter. ETA=0:11:21
[04/07 06:21:01 d2.evaluation.evaluator]: Inference done 226/5000. Dataloading: 0.0296 s/iter. Inference: 0.1122 s/iter. Eval: 0.0000 s/iter. Total: 0.1419 s/iter. ETA=0:11:17

[junjiehe96]

这看起来有些奇怪。不过帧率跟模型运行时具体的硬件和软件环境有关，可以在同一台机器上测一下其他方法以进行公平比较

[supper-zh]

我使用3090GPU，同样的实验设置，FPS都超过40了，确实跟不同的设备性能相关。