orca.data.image.read_parquet support reading as pytorch dataloader

[yangw1234]

API:

orca.data.image.read_parquet(format='tf_dataset|dataloader', input_path=..., transforms=..., config={} **other_kwargs)

• [ ] enable sharding through the config arg. If config has key “size” and “rank”, use them to shard the dataset.
• [ ] enable dataloader in read_parquet

Reference: https://github.com/intel-analytics/analytics-zoo/pull/3956 https://github.com/uber/petastorm#pytorch-api

[yangw1234]

@leonardozcm could you help take a look at this issue?

[leonardozcm]

@leonardozcm could you help take a look at this issue?

Of course, I will have a look.