wkcn edited a comment on issue #12843: MXNet doesn't work with scipy int64 URL: https://github.com/apache/incubator-mxnet/issues/12843#issuecomment-432537269 In SciPy, the dtype `np.int64` is upcasted to `np.longlong` when `csr_matrix` is called. ``` >>> np.int64 == np.longlong False >>> np.int64 <class 'numpy.int64'> >>> np.longlong <class 'numpy.int64'> ``` Please see the link: https://github.com/scipy/scipy/blob/master/scipy/sparse/coo.py#L358 ```python import numpy as np import scipy as sp import mxnet as mx one = np.ones((2, 2), dtype='int64') sp_one = sp.sparse.csr_matrix(one).data print (np.dtype(sp_one.dtype).type) print (id(np.dtype(sp_one.dtype).type)) print (np.dtype(np.int64).type) print (id(np.dtype(np.int64).type)) print (np.int64) print (id(np.int64)) ``` ```bash <type 'numpy.int64'> 139671667680256 <type 'numpy.int64'> 139671667680672 <type 'numpy.int64'> 139671667680672 ```
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services
