ValueError: The input size of each GPU must be 1 in testing mode

[zhang123-sys]

After training GaitGL for 80000 epochs, I test GaitGL by running "CUDA_VISIBLE_DEVICES=0 python -m torch.distributed.launch --master_port 12345 --nproc_per_node=1 opengait/main.py --cfgs ./config/gaitgl/gaitgl.yaml --phase test", but get "ValueError: The input size of each GPU must be 1 in testing mode, but got 4!"

[zhang123-sys]

Traceback (most recent call last):
  File "opengait/main.py", line 71, in <module>
    run_model(cfgs, training)
  File "opengait/main.py", line 56, in run_model
    Model.run_test(model)
  File "/hy-nas/OpenGait/opengait/modeling/base_model.py", line 439, in run_test
    info_dict = model.inference(rank)
  File "/hy-nas/OpenGait/opengait/modeling/base_model.py", line 378, in inference
    retval = self.forward(ipts)
  File "/hy-nas/OpenGait/opengait/modeling/models/gaitgl.py", line 157, in forward
    'The input size of each GPU must be 1 in testing mode, but got {}!'.format(len(labs)))
ValueError: The input size of each GPU must be 1 in testing mode, but got 4!
Transforming:   0%|                                                                                                                                      | 0/5485 [00:00<?, ?it/s]
ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 6432) of binary: /usr/local/miniconda3/envs/OpenGait/bin/python
Traceback (most recent call last):
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/runpy.py", line 193, in _run_module_as_main
    "__main__", mod_spec)
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/runpy.py", line 85, in _run_code
    exec(code, run_globals)
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/launch.py", line 193, in <module>
    main()
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/launch.py", line 189, in main
    launch(args)
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/launch.py", line 174, in launch
    run(args)
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/run.py", line 713, in run
    )(*cmd_args)
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/launcher/api.py", line 131, in __call__
    return launch_agent(self._config, self._entrypoint, list(args))
  File "/usr/local/miniconda3/envs/OpenGait/lib/python3.6/site-packages/torch/distributed/launcher/api.py", line 261, in launch_agent
    failures=result.failures,
torch.distributed.elastic.multiprocessing.errors.ChildFailedError: 
============================================================
opengait/main.py FAILED
------------------------------------------------------------
Failures:
  <NO_OTHER_FAILURES>
------------------------------------------------------------
Root Cause (first observed failure):
[0]:
  time      : 2022-12-17_11:03:35
  host      : Idddbe339d00c011ff
  rank      : 0 (local_rank: 0)
  exitcode  : 1 (pid: 6432)
  error_file: <N/A>
  traceback : To enable traceback see: https://pytorch.org/docs/stable/elastic/errors.html
============================================================

[ChaoFan996]

Hi, set batch size to 1 here, please.

[ChaoFan996]

In OpenGait, we should make batch size equal to the number of used GPUs during testing for models that perform compressing operations over temporal dimension, i.e., GaitGL.

[ChaoFan996]

Moreover, the same to the training phase once we take the unfix-length sequence as input.

[zhang123-sys]

Ok,