Hi Cindy, I don't think cudf supports arrow 7.0.0. Even the master branch is going with 6.0.1 at the moment. https://github.com/rapidsai/cudf/blob/6bcfc104051f926f46467fc55c456a9b012fc4af/conda/environments/cudf_dev_cuda11.5.yml#L20
On Wed, Mar 2, 2022 at 7:21 PM Cindy McMullen <[email protected]> wrote: > pyarrow-7.0.0 > > > /usr/local/cuda/bin/nvcc --version > nvcc: NVIDIA (R) Cuda compiler driver > Copyright (c) 2005-2020 NVIDIA Corporation > Built on Thu_Jun_11_22:26:38_PDT_2020 > Cuda compilation tools, release 11.0, V11.0.194 > Build cuda_11.0_bu.TC445_37.28540450_0 > > > > On Wed, Mar 2, 2022 at 9:09 AM Keith Kraus <[email protected]> > wrote: > >> Hey Cindy, >> >> What versions of cuDF and PyArrow do you have installed? I've typically >> seen this pop up when there's a mismatched version. >> >> -Keith >> >> On Tue, Mar 1, 2022 at 8:28 PM Cindy McMullen <[email protected]> >> wrote: >> >>> Hi - >>> >>> I'm trying to use DGL (Deep Graph Library) DGLDataset API with the >>> RAPIDS cuda DataFrame API. Am getting this error: >>> >>> module 'pyarrow.lib' has no attribute '_CRecordBatchReader' >>> >>> >>> Wonder if you see anything obvious in the stack trace that might help me >>> debug? >>> >>> >>> >>> Here's the full stack trace: >>> >>> >>> <ipython-input-5-81f2b1833437> in __init__(self) 10 """ 11 def >>> __init__(self):---> 12 super(UserSimsSingleFileDataset, >>> self).__init__(name='UserSimsDataset', verbose=False) 13 14 # >>> One quirk of DGLDataset is that process() and __len__ (load, save) are >>> called immediately after super(), >>> /opt/conda/lib/python3.7/site-packages/dgl/data/dgl_dataset.py in >>> __init__(self, name, url, raw_dir, save_dir, hash_key, force_reload, >>> verbose) 91 self._save_dir = save_dir 92 ---> 93 >>> self._load() 94 95 def download(self): >>> /opt/conda/lib/python3.7/site-packages/dgl/data/dgl_dataset.py in >>> _load(self) 176 if not load_flag: 177 >>> self._download()--> 178 self.process() 179 >>> self.save() 180 if self.verbose: >>> <ipython-input-5-81f2b1833437> in process(self) 20 21 import >>> gcsfs---> 22 import cudf 23 self.rows_per_batch = 10000 24 >>> gs = gcsfs.GCSFileSystem() >>> /opt/conda/lib/python3.7/site-packages/cudf/__init__.py in <module> 9 >>> import rmm 10 ---> 11 from cudf import core, datasets, testing 12 >>> from cudf._version import get_versions 13 from cudf.api.extensions >>> import ( >>> /opt/conda/lib/python3.7/site-packages/cudf/core/__init__.py in <module> >>> 1 # Copyright (c) 2018-2020, NVIDIA CORPORATION. 2 ----> 3 from >>> cudf.core import buffer, column, column_accessor, common 4 from >>> cudf.core.buffer import Buffer 5 from cudf.core.dataframe import >>> DataFrame, from_pandas, merge >>> /opt/conda/lib/python3.7/site-packages/cudf/core/column/__init__.py in >>> <module> 1 # Copyright (c) 2020-2021, NVIDIA CORPORATION. 2 ----> >>> 3 from cudf.core.column.categorical import CategoricalColumn 4 from >>> cudf.core.column.column import ( 5 ColumnBase, >>> /opt/conda/lib/python3.7/site-packages/cudf/core/column/categorical.py in >>> <module> 20 21 import cudf---> 22 from cudf import _lib as libcudf >>> 23 from cudf._lib.scalar import as_device_scalar 24 from >>> cudf._lib.transform import bools_to_mask >>> /opt/conda/lib/python3.7/site-packages/cudf/_lib/__init__.py in <module> >>> 2 import numpy as np 3 ----> 4 from . import ( 5 avro, >>> 6 binaryop, >>> cudf/_lib/gpuarrow.pyx in init cudf._lib.gpuarrow() >>> AttributeError: module 'pyarrow.lib' has no attribute '_CRecordBatchReader' >>> >>> -- Niranda Perera https://niranda.dev/ @n1r44 <https://twitter.com/N1R44>
