AttributeError: 'DetectFeatLmdb' object has no attribute 'env'

when I run python train_re.py --config config/train-refcoco-base-1gpu.json --output_dir $RE_EXP

I just get this result.

03/11/2021 11:44:06 - INFO - __main__ - Loading Train Dataset /txt_db/refcoco_train.db, /img_db/re_coco_gt
/img_db/re_coco_gt
/txt_db/refcoco_train.db
Traceback (most recent call last):
File "train_re.py", line 461, in
main(args)
File "train_re.py", line 129, in main
ReDataset, re_collate, opts)
File "train_re.py", line 46, in create_dataloader
num_bb, opts.compressed_db)
File "/src/data/data.py", line 75, in __init__
readahead=not _check_distributed())
lmdb.Error: /img_db/re_coco_gt/feat_numbb100: No such file or directory
Exception ignored in: >
Traceback (most recent call last):
File "/src/data/data.py", line 98, in __del__
self.env.close()
AttributeError: 'DetectFeatLmdb' object has no attribute 'env'

Does anyone know this?

Reproduce the command from train_re.py and inspect create_dataloader around line 46, then trace DetectFeatLmdb initialization and cleanup in data/data.py around lines 75 and 98. Check the expected dataset paths against the configured /img_db/re_coco_gt location; done means the command loads the dataset without the missing-path failure or the secondary env AttributeError.

Written by the indexing model from the issue text.