kjunelee/MetaOptNet

qpth Error: Cannot perform LU factorization on Q.

Ze-Yang opened this issue · 4 comments

Ze-Yang commented

Pytorch 1.0.1. Still have this problem with RTX2080ti.

WHQ1111 commented

Pytorch 1.3.1 + Python 3.6 have same problem.

kjunelee commented

I am not sure why this happens. FYI, I am able to run the code with qpth 0.0.13.

cryu854 commented

Upgrade the version of qpth might be work.
my packages: Python3.7.6 + Pytorch 1.2.0 + qpth 0.015

  • 👍2
psandovalsegura commented

The suggestion by @cryu854 seems to have worked for me (using torch==1.2.0 and qpth==0.0.15). The script runs now, but there are a number of other warnings like:

/pytorch/torch/csrc/autograd/python_function.cpp:638: UserWarning: Legacy autograd function with non-static forward method is deprecated and will be removed in 1.3. Please use new-style autograd function with static forward method. (Example: https://pytorch.org/docs/stable/autograd.html#torch.autograd.Function)
/pytorch/aten/src/ATen/native/IndexingUtils.h:20: UserWarning: indexing with dtype torch.uint8 is now deprecated, please use a dtype torch.bool instead.
/pytorch/aten/src/ATen/native/BatchLinearAlgebra.cpp:1162: UserWarning: Passing RHS tensor with number of dimensions = 2 is deprecated, and will be removed in the next release. Please unsqueeze the last dimension to obtain an RHS tensor with number of right hand sides = 1

Not sure if any of these impact the correctness of training.