Running benchmark.py reports an error

Dear developer, I want to calculate the inference time of Faster R-CNN on my computer, so I am going to use the benchmark.py file. The command entered is as follows. python -m torch.distributed.launch --nproc_per_node=1 --master_port=29500 tools/analysis_tools/benchmark.py configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py work_dirs/faster_rcnn_r50_fpn_1x_coco/epoch_100.pth --launcher pytorch --fuse-conv-bn However, I encountered the following error report, how can I solve it? My computer system is Win10, GPU is Quadro P5000 single card, CUDA is 10.2. Thank you.

Traceback (most recent call last):
  File "tools/analysis_tools/benchmark.py", line 143, in <module>
    main()
  File "tools/analysis_tools/benchmark.py", line 136, in main
    init_dist(args.launcher, **cfg.dist_params)
  File "f:\pycharmprojects\mmdetection_env\mmcv\mmcv\runner\dist_utils.py", line 20, in init_dist
    _init_dist_pytorch(backend, **kwargs)
  File "f:\pycharmprojects\mmdetection_env\mmcv\mmcv\runner\dist_utils.py", line 34, in _init_dist_pytorch
    dist.init_process_group(backend=backend, **kwargs)
  File "D:\Anaconda3\envs\mmdetection_env\lib\site-packages\torch\distributed\distributed_c10d.py", line 433, in init_process_group
    rendezvous_iterator = rendezvous(
  File "D:\Anaconda3\envs\mmdetection_env\lib\site-packages\torch\distributed\rendezvous.py", line 82, in rendezvous
    raise RuntimeError("No rendezvous handler for {}://".format(result.scheme))
RuntimeError: No rendezvous handler for env://
Traceback (most recent call last):
  File "D:\Anaconda3\envs\mmdetection_env\lib\runpy.py", line 194, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "D:\Anaconda3\envs\mmdetection_env\lib\runpy.py", line 87, in _run_code
    exec(code, run_globals)
  File "D:\Anaconda3\envs\mmdetection_env\lib\site-packages\torch\distributed\launch.py", line 260, in <module>
    main()
  File "D:\Anaconda3\envs\mmdetection_env\lib\site-packages\torch\distributed\launch.py", line 255, in main
    raise subprocess.CalledProcessError(returncode=process.returncode,
subprocess.CalledProcessError: Command '['D:\\Anaconda3\\envs\\mmdetection_env\\python.exe', '-u', 'tools/analysis_tools/benchmark.py', '--local_rank=0', 'configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py', 'work_dirs/faster_rcnn_r50_
fpn_1x_coco/epoch_100.pth', '--launcher', 'pytorch', '--fuse-conv-bn']' returned non-zero exit status 1.

[shinya7y]

Does old code with MMDataParallel before https://github.com/open-mmlab/mmdetection/pull/5265 work? https://github.com/open-mmlab/mmdetection/blob/2ca1323cf2754c88a57fe645163c0bf2cd79a3e0/tools/analysis_tools/benchmark.py

[zxsitu]

@shinya7y Thanks for your prompt advice, I tested it and it ran successfully!!! My computer (Quadro P5000) ran Faster RCNN Resnet50 with a result of 15.8 FPS. I believe this result should be quite reasonable.

(mmdetection_env) F:\PycharmProjects\mmdetection_env\mmdetection>python tools/analysis_tools/benchmark.py configs/faster_rcnn/faster_rcnn_r50_fpn_1x_coco.py work_dirs/faster_rcnn_r50_fpn_1x_coco/epoch_100.pth --fuse-conv-bn
D:\Anaconda3\envs\mmdetection_env\lib\site-packages\mmdet-2.15.1-py3.8.egg\mmdet\core\anchor\builder.py:16: UserWarning: ``build_anchor_generator`` would be deprecated soon, please use ``build_prior_generator``
  warnings.warn(
Use load_from_local loader
D:\Anaconda3\envs\mmdetection_env\lib\site-packages\mmdet-2.15.1-py3.8.egg\mmdet\core\anchor\anchor_generator.py:324: UserWarning: ``grid_anchors`` would be deprecated soon. Please use ``grid_priors``
  warnings.warn('``grid_anchors`` would be deprecated soon. '
D:\Anaconda3\envs\mmdetection_env\lib\site-packages\mmdet-2.15.1-py3.8.egg\mmdet\core\anchor\anchor_generator.py:360: UserWarning: ``single_level_grid_anchors`` would be deprecated soon. Please use ``single_level_grid_priors``
  warnings.warn(
Done image [50 / 2000], fps: 15.7 img / s
Done image [100/ 2000], fps: 15.7 img / s
Done image [150/ 2000], fps: 15.7 img / s
Done image [200/ 2000], fps: 15.8 img / s