resnet_serialize_test failure on weight type

[jabowery]

The attached file is terminal output from a complete install (under Ubuntu 18.04 (bionic) with CUDA 10.2) as per README:

pip install -r requirements.txt
pip install -r requirements-dev.txt
python setup.py develop

At the end is the error, the complete error message:

E RuntimeError: Input type (torch.FloatTensor) and weight type (torch.cuda.FloatTensor) should be the same

nupicresearchsetuperror.txt

[jabowery]

The same test error arose after the alternate installation procedure:

conda env create

However, before I could get that installation procedure to run successfully, I had to do 4 things:

1. Edit nupic.research/environment.yml to remove the version spec from grpcio>=1.24.3.
2. Create and cd ${HOME}/nta
3. git clone https://github.com/numenta/nupic.torch.git
4. git clone https://github.com/numenta/nupic.tensorflow.git

[jabowery]

Apparently, the input needs to be pushed to CUDA as well as the model. See:

https://discuss.pytorch.org/t/input-type-torch-floattensor-and-weight-type-torch-cuda-floattensor-should-be-the-same/48633