Added a ADE20K and COCO2017 data conversion scripts

[karan6181]

Description

• Added a ADE20K and COCO2017 data conversion scripts
• Currently, it converts into mds format.
• User can run the script as a standalone script to convert their RAW dataset into mds format.

Linting

$ isort <file_name>
$ yapf -i -vv -p <file_name>
$ pyright

Testing

• Downloaded the raw ADE20K dataset locally and ran the ade20k.py script locally to generate the sharded mds file. The snapshot of structure is as shown below

.
├── train
│   ├── index.json
│   ├── shard.00000.mds
│   ├── shard.00001.mds
│   ├── shard.00002.mds
│   ├── shard.00003.mds
│   ├── shard.00004.mds
│   ├── shard.00005.mds
│   ├── shard.00006.mds
|    ............
│   ├── shard.00208.mds
│   ├── shard.00209.mds
│   └── shard.00210.mds
└── val
    ├── index.json
    ├── shard.00000.mds
    ├── shard.00001.mds
    ├── shard.00002.mds
     ............
    ├── shard.00018.mds
    ├── shard.00019.mds
    ├── shard.00020.mds
    └── shard.00021.mds

2 directories, 235 files

• Downloaded the raw MSCOCO-2017 dataset locally and ran the coco.py script locally to generate the sharded mds file. The curtailed snapshot of structure is as shown below

.
├── train
│   ├── index.json
│   ├── shard.00000.mds
│   ├── shard.00001.mds
│   ├── shard.00002.mds
│   ├── shard.00003.mds
│   ├── shard.00004.mds
│   ├── shard.00005.mds
│   ├── shard.00006.mds
│   ├── shard.00007.mds.
│   ├── shard.00008.mds
│   ├── shard.00009.mds
│   ├── shard.00010.mds
|    .........
│   ├── shard.00572.mds
│   └── shard.00573.mds
└── val
    ├── index.json
    ├── shard.00000.mds
    ├── shard.00001.mds
    ├── shard.00002.mds
    ├── shard.00003.mds
     .........
    ├── shard.00022.mds
    ├── shard.00023.mds
    └── shard.00024.mds

2 directories, 601 files