Ceph support for NuScenes evaluation.

[Lotus-95]

I have NuScenes data files on ceph, so I config as https://mmdetection3d.readthedocs.io/en/latest/tutorials/backends_support.html

It works when model training, but when evaluating, it doesn't work.

Evaluating bboxes of pts_bbox Traceback (most recent call last): File "tools/train.py", line 257, in main() File "tools/train.py", line 253, in main meta=meta) File "/mnt/lustre/{username}/workspace/mmdetection3d/mmdet3d/apis/train.py", line 35, in train_model meta=meta) File "/mnt/lustre/{username}/workspace/mmdetection/mmdet/apis/train.py", line 244, in train_detector runner.run(data_loaders, cfg.workflow) File "/mnt/lustre/{username}/workspace/mmcv/mmcv/runner/epoch_based_runner.py", line 127, in run epoch_runner(data_loaders[i], **kwargs) File "/mnt/lustre/{username}/workspace/mmcv/mmcv/runner/epoch_based_runner.py", line 54, in train self.call_hook('after_train_epoch') File "/mnt/lustre/{username}/workspace/mmcv/mmcv/runner/base_runner.py", line 307, in call_hook getattr(hook, fn_name)(self) File "/mnt/lustre/{username}/workspace/mmcv/mmcv/runner/hooks/evaluation.py", line 267, in after_train_epoch self._do_evaluate(runner) File "/mnt/lustre/{username}/workspace/mmdetection/mmdet/core/evaluation/eval_hooks.py", line 125, in _do_evaluate key_score = self.evaluate(runner, results) File "/mnt/lustre/{username}/workspace/mmcv/mmcv/runner/hooks/evaluation.py", line 362, in evaluate results, logger=runner.logger, **self.eval_kwargs) File "/mnt/lustre/{username}/workspace/mmdetection3d/mmdet3d/datasets/nuscenes_dataset.py", line 505, in evaluate ret_dict = self._evaluate_single(result_files[name]) File "/mnt/lustre/{username}/workspace/mmdetection3d/mmdet3d/datasets/nuscenes_dataset.py", line 391, in _evaluate_single version=self.version, dataroot=self.data_root, verbose=False) File "/mnt/lustre/share_data/{username}/bev/lib/python3.6/site-packages/nuscenes/nuscenes.py", line 62, in init assert osp.exists(self.table_root), 'Database version not found: {}'.format(self.table_root) AssertionError: Database version not found: s3://nuscenes/v1.0-trainval

[ZCMax]

For now, since the nuscenes evaluation relies on the official api, which does not support ceph. If you want to do nuscenes evaluatoin, you need to prepare maps and v1.0-trainval folder under your local data/nuscenes directory.

[ZCMax]

We'll consider to support ceph for nuscenes evaluation in th future.