Cannot read from parquet files that contain binary

🐛 Describe the bug

I create some parquet files with the following:

def save_tensor(t):
    buf = io.BytesIO()
    th.save(t, buf)
    return buf.getvalue()

    for idx, batch in enumerate(tqdm(dl, total=total_files / args.batch_size)):
        df = pa.table({"img": [save_tensor(item) for item in batch]})
        pq.write_table(df, out_dir / f"{idx}.parquet")

when I try to read the parquet files, with the following pipe:

    datapipe = dp.iter.FSSpecFileLister(dir)
    datapipe = datapipe.load_parquet_as_df()

I get the error:

NotImplementedError: Unsupported Arrow type: binary
This exception is thrown by iter of ParquetDFLoaderIterDataPipe(columns=None, device='', dtype=None, > source_dp=FSSpecFileListerIterDataPipe, use_threads=False)

Versions

Collecting environment information...
PyTorch version: 1.12.0+cu102
Is debug build: False
CUDA used to build PyTorch: 10.2
ROCM used to build PyTorch: N/A

OS: Ubuntu 22.04 LTS (x86_64)
GCC version: (Ubuntu 11.2.0-19ubuntu1) 11.2.0
Clang version: Could not collect
CMake version: Could not collect
Libc version: glibc-2.35

Python version: 3.9.12 (main, Jun 1 2022, 11:38:51) [GCC 7.5.0] (64-bit runtime)
Python platform: Linux-5.15.0-41-generic-x86_64-with-glibc2.35
Is CUDA available: True
CUDA runtime version: Could not collect
GPU models and configuration: GPU 0: NVIDIA GeForce GTX 1080 Ti
Nvidia driver version: 515.48.07
cuDNN version: Could not collect
HIP runtime version: N/A
MIOpen runtime version: N/A
Is XNNPACK available: True

Versions of relevant libraries:
[pip3] mypy-extensions==0.4.3
[pip3] numpy==1.21.4
[pip3] pytorch-ranger==0.1.1
[pip3] torch==1.12.0
[pip3] torch-optimizer==0.1.0
[pip3] torcharrow==0.1.0
[pip3] torchdata==0.4.0
[pip3] torchmetrics==0.7.3
[pip3] torchvision==0.13.0
[conda] cudatoolkit 11.7.0 hd8887f6_10 conda-forge
[conda] numpy 1.21.4 pypi_0 pypi
[conda] pytorch-ranger 0.1.1 pypi_0 pypi
[conda] torch 1.12.0 pypi_0 pypi
[conda] torch-optimizer 0.1.0 pypi_0 pypi
[conda] torcharrow 0.1.0 pypi_0 pypi
[conda] torchdata 0.4.0 pypi_0 pypi
[conda] torchmetrics 0.7.3 pypi_0 pypi
[conda] torchvision 0.13.0 pypi_0 pypi

Start at datapipe.load_parquet_as_df() and the ParquetDFLoaderIterDataPipe path shown in the error, then trace how Arrow binary columns are converted. Reproduce the issue with the provided save_tensor, pa.table, and pq.write_table example; done means the generated parquet file loads without the Unsupported Arrow type: binary error. The payload names no source file or test, so locate those first.