Converting sparse matrix errors

Am seeing the following error when trying to convert a csc_matrix to a coo_matrix. Though also see some variation of this when trying to go to a csr_matrix from a csc_matrix and other similar conversions.

• Conditions (you can just paste the output of python -c 'import cupy; cupy.show_config()')

CuPy Version          : 7.2.0
CUDA Root             : /datasets/jkirkham/miniconda/envs/rapids13dev
CUDA Build Version    : 10010
CUDA Driver Version   : 10020
CUDA Runtime Version  : 10010
cuBLAS Version        : 10201
cuFFT Version         : 10101
cuRAND Version        : 10101
cuSOLVER Version      : (10, 2, 0)
cuSPARSE Version      : 10300
NVRTC Version         : (10, 1)
cuDNN Build Version   : 7605
cuDNN Version         : 7605
NCCL Build Version    : 2406
NCCL Runtime Version  : 2507

• Code to reproduce

In [1]: import cupy                                                             

In [2]: a = cupy.array([[0, 1, 0], 
   ...:                 [2, 0, 3], 
   ...:                 [0, 4, 0]], dtype=float)                                

In [3]: a2 = cupy.sparse.csc_matrix(a)                                          

In [4]: a3 = a2.tocoo()                                                         

• Error messages, stack traces, or logs

---------------------------------------------------------------------------
CUDADriverError                           Traceback (most recent call last)
<ipython-input-4-14bf9613c858> in <module>
----> 1 a3 = a2.tocoo()

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupyx/scipy/sparse/csc.py in tocoo(self, copy)
    179 
    180         """
--> 181         return self.T.tocoo(copy).T
    182 
    183     def tocsc(self, copy=None):

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupyx/scipy/sparse/csr.py in tocoo(self, copy)
    239             indices = self.indices
    240 
--> 241         return cusparse.csr2coo(self, data, indices)
    242 
    243     def tocsc(self, copy=False):

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupy/cusparse.py in csr2coo(x, data, indices)
    631     # data and indices did not need to be copied already
    632     return cupyx.scipy.sparse.coo_matrix(
--> 633         (data, (row, indices)), shape=x.shape)
    634 
    635 

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupyx/scipy/sparse/coo.py in __init__(self, arg1, shape, dtype, copy)
    128 
    129         if len(data) > 0:
--> 130             if row.max() >= shape[0]:
    131                 raise ValueError('row index exceeds matrix dimensions')
    132             if col.max() >= shape[1]:

cupy/core/core.pyx in cupy.core.core.ndarray.__richcmp__()

cupy/core/_kernel.pyx in cupy.core._kernel.ufunc.__call__()

cupy/core/_kernel.pyx in cupy.core._kernel.ufunc._get_ufunc_kernel()

cupy/core/_kernel.pyx in cupy.core._kernel._get_ufunc_kernel()

cupy/core/_kernel.pyx in cupy.core._kernel._get_simple_elementwise_kernel()

cupy/core/carray.pxi in cupy.core.core.compile_with_cache()

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupy/cuda/compiler.py in compile_with_cache(source, options, arch, cache_dir, extra_source, backend)
    285     else:
    286         return _compile_with_cache_cuda(source, options, arch, cache_dir,
--> 287                                         extra_source, backend)
    288 
    289 

/datasets/jkirkham/miniconda/envs/rapids13dev/lib/python3.7/site-packages/cupy/cuda/compiler.py in _compile_with_cache_cuda(source, options, arch, cache_dir, extra_source, backend)
    333             cubin_hash = six.b(hashlib.md5(cubin).hexdigest())
    334             if hash == cubin_hash:
--> 335                 mod.load(cubin)
    336                 return mod
    337 

cupy/cuda/function.pyx in cupy.cuda.function.Module.load()

cupy/cuda/function.pyx in cupy.cuda.function.Module.load()

cupy/cuda/driver.pyx in cupy.cuda.driver.moduleLoadData()

cupy/cuda/driver.pyx in cupy.cuda.driver.check_status()

CUDADriverError: CUDA_ERROR_ILLEGAL_ADDRESS: an illegal memory access was encountered

[leofang]

Confirmed this is fixed in the updated conda package (conda-forge/cupy-feedstock#39).

[jakirkham]

Sure if it is helpful to have open. No objections. Reopening…

Thanks Leo 🙂