Where is the sensorimotor code?

Engineering
21

@mccall.ryan @thanh-binh.to I will be able to help within the next few months. I will be running SMI code for upcoming HTM School episodes. I’ll make sure there are clear instructions for installation and running some examples. But it will not be right away.

1 Like
22

@rhyolight thanks, but i really need 2 things: 1) how is the python path in your laptop? 2) in which py-file is ExtendedTemporalMemory implemented? In nupic/bindings/experimental.py ?
@mrcslws @rhyolight can you give me those information?

23

We can do some tricks for solving this problem like copying the file “experimental.py” from htmresearch-core into htmresearch folder and modify swig binding scripts. But trick is trick.
@mrcslws what is your opinion?

24

@thanh-binh.to, @mccall.ryan We are currently working on making htmresearch easier to install, similar to nupic where you would be able to simply pip install htmresearch, but for now you can try the steps we use internally to build and install htmresearch on our dev laptops:

  1. Build nupic.core
pip install --user numpy==1.12.1 pytest==3.0.7 pytest-cov==2.5.0 pytest-xdist==1.16.0 pycapnp==0.5.12
mkdir -p ~/nta/nupic.core/build/scripts_release
cd ~/nta/nupic.core/build/scripts_release
cmake ../.. -DCMAKE_INSTALL_PREFIX=../release -DCMAKE_BUILD_TYPE=Release -DNUPIC_IWYU=OFF -DNUPIC_TOGGLE_INSTALL=ON -DPY_EXTENSIONS_DIR=~/nta/nupic.core/bindings/py/src/nupic/bindings
make -j6
make install
  1. Instal nupic.core
cd ~/nta/nupic.core
pip install --user numpy==1.12.1 pytest==3.0.7 pytest-cov==2.5.0 pytest-xdist==1.16.0 pycapnp==0.5.12
ARCHFLAGS=-arch x86_64 pip install --user -e .
  1. Install nupic
cd ~/nta/nupic
pip install --user asteval==0.9.1 coverage==3.7.1 mock==1.0.1 ordereddict==1.1 psutil==1.0.1 pytest==3.0.7 pytest-cov==2.5.0 pytest-xdist==1.16.0 python-dateutil==2.1 PyYAML==3.10 unittest2==0.5.1 validictory==0.9.1 PyMySQL==0.6.2 DBUtils==1.1 pyproj==1.9.3 prettytable==0.7.2  numpy==1.12.1
ARCHFLAGS=-arch x86_64 pip install --user -e .
  1. Build htmresearch-core
mkdir -p ~/nta/htmresearch-core/build/scripts_release
cd ~/nta/htmresearch-core/build/scripts_release
cmake ../.. -DCMAKE_INSTALL_PREFIX=../release -DCMAKE_BUILD_TYPE=Release -DNUPIC_IWYU=OFF -DLOCAL_NUPIC_CORE_INSTALL_DIR=~/nta/nupic.core/build/release -DPY_EXTENSIONS_DIR=~/nta/htmresearch-core/bindings/py/src/htmresearch_core
make -j6
make install
  1. Install htmresearch-core
cd ~/nta/htmresearch
pip install --user numpy==1.12.1 pytest==3.0.7 pytest-cov==2.5.0 pytest-xdist==1.16.0 pycapnp==0.5.12
ARCHFLAGS=-arch x86_64 pip install --user -e .
  1. Install htmresearch
cd ~/nta/htmresearch-core
pip install --user matplotlib pandas plotly>=1.12.10 simplejson tabulate tqdm prettytable plyfile scikit-learn
ARCHFLAGS=-arch x86_64 pip install --user -e .

Let me know if it works for you

1 Like
25

@lscheinkman thanks for your guide. I am 98% sure that i installed in the same ways/steps like you. As i mentioned last post, the file “experimental.py” is missing in htmresearch and even by copying it manuelly from htmresearch-core i have an error by swig bindings helper where the module “_experimental” is not found.

26

@thanh-binh.to experimental.py is a file generated by cmake/swig, see https://github.com/numenta/htmresearch-core/blob/master/src/CMakeLists.txt#L543

You should not copy it from other locations as it is just a build artifact. Try the above instructions using the latest version from github and make sure to clean your repository before running the commands.

1 Like
27

@lscheinkman perfect. it works now using your installation guide. Best thanks for your support!

1 Like
28

By testing the SMI codes, exactly in project l2_pooling, I found that even though both the location and feature are input into L2L4 experiment, but only the location information is encoded using CoordinateEncoder. I can not find out if and how the “feature” will be used? without encoder? Can anyone help me?

29

Moved from #htm-theory to #htm-hackers.

@thanh-binh.to Please realize this code is still alpha and not ready for real use. Don’t forget the disclaimer: What you should understand about this repository.

30

Which file in the repo should I start with to study the latest pooling algorithm used in the output layers?

1 Like
31

@rhyolight I totally agree with you about using this repository. I just want only to learn more about those experiments …and try to discuss …

1 Like