[Feature] Support Flickr30k Retrieval dataset

[InvincibleWyq]

Modification

Support Flickr30k dataset.

A example config for BLIP zero-shot retrieval is provided. It matches the result provided in the paper.

A example config for BLIP zero-shot captioning is provided. Its result is reasonable.

Example command:

sh tools/slurm_test.sh <patrition> <jobname> configs/blip/blip-base_8xb32_retrieval_flickr30k.py /mnt/petrelfs/share_data/wangyiqin/blip_ckpt/blip-base_3rdparty_coco-retrieval_20230419-a1804d2c.pth
sh tools/slurm_test.sh <patrition> <jobname> configs/blip/blip-base_8xb32_caption_flickr30k.py /mnt/petrelfs/share_data/wangyiqin/blip_ckpt/blip-base_3rdparty_coco-caption_20230419-a5b71af3.pth

Local Path
Images and annotations at: /mnt/petrelfs/share_data/liuyuan/data/mm/flickr30k
Ground truth annotations for evaluation at: /mnt/petrelfs/share_data/wangyiqin/flickr30k

[codecov]

Codecov Report

Patch coverage: 36.98% and project coverage change: -2.67 ⚠️

Comparison is base (f9dcae2) 68.16% compared to head (e1f4b2e) 65.50%.

❗ Current head e1f4b2e differs from pull request most recent head bd45198. Consider uploading reports for the commit bd45198 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##              dev    #1625      +/-   ##
==========================================
- Coverage   68.16%   65.50%   -2.67%     
==========================================
  Files         295      328      +33     
  Lines       23372    25561    +2189     
  Branches     3713     4075     +362     
==========================================
+ Hits        15932    16744     +812     
- Misses       6880     8207    +1327     
- Partials      560      610      +50     

Flag	Coverage Δ
unittests	65.50% <36.98%> (-2.67%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmpretrain/apis/image_retrieval.py	21.42% <ø> (ø)
mmpretrain/datasets/__init__.py	60.46% <0.00%> (-13.83%)	⬇️
mmpretrain/datasets/flickr30k_caption.py	0.00% <0.00%> (ø)
mmpretrain/datasets/flickr30k_retrieval.py	0.00% <0.00%> (ø)
mmpretrain/datasets/gqa_dataset.py	0.00% <0.00%> (ø)
mmpretrain/datasets/nocaps.py	0.00% <0.00%> (ø)
mmpretrain/datasets/ocr_vqa.py	0.00% <0.00%> (ø)
mmpretrain/datasets/scienceqa.py	0.00% <0.00%> (ø)
mmpretrain/datasets/textvqa.py	0.00% <0.00%> (ø)
mmpretrain/datasets/transforms/__init__.py	100.00% <ø> (ø)
... and 55 more

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[fangyixiao18]

Suggested change

	evaluation, refer to refer
	evaluation, please refer to