🐛 Describe the bug

Pinning collated batches of size 1 of strings yields the string in a list instead of the string itself.
E.g. ['hello'] -> ['h', 'e', 'l', 'l', 'o']

from torchdata.datapipes.iter import IterableWrapper

dp = IterableWrapper(["hello", "world"])
dp = dp.batch(1)
dp = dp.collate()
dp = dp.pin_memory()

print(next(iter(dp)))

['h', 'e', 'l', 'l', 'o']

Seems to be because in the pin_memory_fn on line 30 type(data)(*pinned_data) calls list on the string itself. For batch sizes > 1 this leads to an error which just returns pinned_data.
Should this line be type(data)(pinned_data)?

Versions

torchdata.version=='0.7.0a0+cacf355'