[Tip] Error in learning Tokenizer

[soonchanpark]

Hi, i'm working with the code you shared. Thank you for your work.

In my evironment settings, I have met an error below when executing training code of Tokenizer.

Traceback (most recent call last):
  File "tools/train.py", line 175, in <module>
    main()
  File "tools/train.py", line 171, in main
    meta=meta)
  File "/opt/conda/lib/python3.7/site-packages/mmpose/apis/train.py", line 213, in train_model
    runner.run(data_loaders, cfg.workflow, cfg.total_epochs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 136, in run
    epoch_runner(data_loaders[i], **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 53, in train
    self.run_iter(data_batch, train_mode=True, **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/runner/epoch_based_runner.py", line 32, in run_iter
    **kwargs)
  File "/opt/conda/lib/python3.7/site-packages/mmcv/parallel/data_parallel.py", line 77, in train_step
    return self.module.train_step(*inputs[0], **kwargs[0])
  File "/opt/conda/lib/python3.7/site-packages/mmpose/models/detectors/base.py", line 106, in train_step
    loss, log_vars = self._parse_losses(losses)
  File "/opt/conda/lib/python3.7/site-packages/mmpose/models/detectors/base.py", line 51, in _parse_losses
    for loss_name, loss_value in losses.items():
AttributeError: 'tuple' object has no attribute 'items'

In conclusion, compatilbility between pct_detector.py and mmpose==0.29.0 has problem. To solve the error, i modify the code in mmpose/models/detectors/base.py, function train_step, "losses = self.forward(**data_batch, return_loss=True) to run train forward which output losses.