chrischoy/gesvd

Pytorch extension for Singular Value Decompostion (SVD) with LAPACK gesvd backend

gesvd's Issues

module not found

hi Chris,

Under a virtual env, I successfully install gesvd via python setup.py install. I can find it in pip list:

(torch) shenhuan@pf-pc30:.../MinkowskiEngine$ cd ..
(torch) shenhuan@pf-pc30:.../shengyu$ pip show gesvd
Name: gesvd
Version: 0.1
Summary: UNKNOWN
Home-page: UNKNOWN
Author: UNKNOWN
Author-email: UNKNOWN
License: UNKNOWN
Location: /scratch/shengyu/torch/lib/python3.6/site-packages/gesvd-0.1-py3.6-linux-x86_64.egg
Requires: 
Required-by: 

But when importing it, it reports module not found:

(torch) shenhuan@pf-pc30:.../shengyu$ python3
Python 3.6.9 (default, Nov  7 2019, 10:44:02) 
[GCC 8.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import gesvd
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'gesvd'

I install MinkowskiEngine in the same way and can import it, here is pip show message:

(torch) shenhuan@pf-pc30:.../shengyu$ pip show MinkowskiEngine
Name: MinkowskiEngine
Version: 0.4.2
Summary: a convolutional neural network library for sparse tensors
Home-page: https://github.com/StanfordVL/MinkowskiEngine
Author: Christopher Choy
Author-email: [email protected]
License: UNKNOWN
Location: /scratch/shengyu/torch/lib/python3.6/site-packages/MinkowskiEngine-0.4.2-py3.6-linux-x86_64.egg
Requires: numpy, torch
Required-by:

Do you have any clues where might be wrong?

The package doesn't get installed correctly

I followed the instruction to install the package but I got this error

.../anaconda3/envs/py/lib/python3.7/site-packages/gesvd-0.1-py3.7-linux-x86_64.egg/gesvd_cpp.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN6caffe26detail37_typeMetaDataInstance_preallocated_32E

is it stable when doing backprop?

is it stable when doing backprop?