lmdb.Error: /userhome/UNITER/UNITER/img/coco_train2014/all: No such file or directory

(base) root@e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0:/userhome/UNITER/UNITER# rm -rf output && python pretrain.py --config config/pretrain-alldata-base-8gpu.json --output_dir /userhome/UNITER/UNITER/output/
--------------------------------------------------------------------------
[[48719,1],0]: A high-performance Open MPI point-to-point messaging module
was unable to find any relevant network interfaces:

Module: OpenFabrics (openib)
Host: e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0

Another transport will be used instead, although this may result in
lower performance.

NOTE: You can disable this warning by setting the MCA parameter
btl_base_warn_component_unused to 0.
--------------------------------------------------------------------------
07/15/2021 08:44:23 - INFO - __main__ - device: cuda:0 n_gpu: 1, rank: 0, 16-bits training: True
07/15/2021 08:44:24 - INFO - __main__ - Waiting on git info....
07/15/2021 08:44:24 - INFO - __main__ - Git branch: master
07/15/2021 08:44:24 - INFO - __main__ - Git SHA: 1dbfa62bb8ddb1f2b11d20775ce324c5e45a3ab4
0%| | 0/200000 [00:00
main(args)
File "pretrain.py", line 206, in main
opts.train_datasets, True, opts)
File "pretrain.py", line 125, in create_dataloaders
img_db = [all_img_dbs[path] for path in dset['img']]
File "pretrain.py", line 125, in
img_db = [all_img_dbs[path] for path in dset['img']]
File "/userhome/UNITER/UNITER/data/data.py", line 311, in __getitem__
self.min_bb, self.num_bb, self.compress)
File "/userhome/UNITER/UNITER/data/data.py", line 75, in __init__
readahead=not _check_distributed())
lmdb.Error: /userhome/UNITER/UNITER/img/coco_train2014/all: No such file or directory
Exception ignored in: >
Traceback (most recent call last):
File "/userhome/UNITER/UNITER/data/data.py", line 98, in __del__
self.env.close()
AttributeError: 'DetectFeatLmdb' object has no attribute 'env'

(base) root@e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0:/userhome/UNITER/UNITER#
(base) root@e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0:/userhome/UNITER/UNITER#
(base) root@e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0:/userhome/UNITER/UNITER#
(base) root@e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0:/userhome/UNITER/UNITER# rm -rf output && python pretrain.py --config config/pretrain-indomain-base-8gpu.json --output_dir /userhome/UNITER/UNITER/output/

--------------------------------------------------------------------------
[[47594,1],0]: A high-performance Open MPI point-to-point messaging module
was unable to find any relevant network interfaces:

Module: OpenFabrics (openib)
Host: e3ff8e900e504011eb09ab60a9de5c5cd03e-wangx3189-0

Another transport will be used instead, although this may result in
lower performance.

NOTE: You can disable this warning by setting the MCA parameter
btl_base_warn_component_unused to 0.
--------------------------------------------------------------------------
07/15/2021 08:49:38 - INFO - __main__ - device: cuda:0 n_gpu: 1, rank: 0, 16-bits training: True
07/15/2021 08:49:38 - INFO - __main__ - Waiting on git info....
07/15/2021 08:49:38 - INFO - __main__ - Git branch: master
07/15/2021 08:49:38 - INFO - __main__ - Git SHA: 1dbfa62bb8ddb1f2b11d20775ce324c5e45a3ab4
0%| | 0/200000 [00:00
main(args)
File "pretrain.py", line 206, in main
opts.train_datasets, True, opts)
File "pretrain.py", line 125, in create_dataloaders
img_db = [all_img_dbs[path] for path in dset['img']]
File "pretrain.py", line 125, in
img_db = [all_img_dbs[path] for path in dset['img']]
File "/userhome/UNITER/UNITER/data/data.py", line 311, in __getitem__
self.min_bb, self.num_bb, self.compress)
File "/userhome/UNITER/UNITER/data/data.py", line 75, in __init__
readahead=not _check_distributed())
**lmdb.Error: /userhome/UNITER/UNITER/img/coco_train2014/all: No such file or directory**
Exception ignored in: >
Traceback (most recent call last):
File "/userhome/UNITER/UNITER/data/data.py", line 98, in __del__
self.env.close()
**AttributeError: 'DetectFeatLmdb' object has no attribute 'env'**

============================================
Many bugs in this code. Any plan for the update of this code?

Start with the dataset-loading path in pretrain.py, especially create_dataloaders, and then inspect data/data.py around DetectFeatLmdb and its initialization cleanup. Reproduce the two commands from the issue with the referenced configuration files and trace how the coco_train2014 path is selected. Done should mean the reported missing-path failure and subsequent cleanup error have an explicitly defined resolution.

Written by the indexing model from the issue text.