Skip to main content

A high performance implementation of Numenta's HTM algorithms

Project description

PyEtaler

This is the offical Python binding for Etaler. PyEtaler generates Python binging using cppyy and adds additional feature on top of the automatically generated bindings.

Note: As of now, installing cppyy (thus PyEtaler) will cause ROOT to fail to load due to dependency clash.

Installation

Note: You must have Etaler and cppyy installed globally before building the binding. Note: Since the binding is generated to load the actual Etaler installation. You'll need to re-compile the binding everytime Etaler is updated.

If you are building from source (building via directly interacting with the generator).

pip install cppyy # must installed globally
python3 genbinding.py
cp *.so etaler/
cp *.pcm etaler/
# Then copy the resulting files into your package directory

Locally build via PIP

pip3 install .

Alternativelly you can install it directly via PyPI.

pip install pyetaler

Usage

After installation, you can use Etaler from python. The API is exactly like it is in C++.

>>> from etaler import et
>>> et.ones([2, 2])
{{ 1, 1}, 
 { 1, 1}}
>>> sp = et.SpatialPooler([128], [32])
>>> x = et.encoder.scalar(0.1, 0, 1, 128, 12)
>>> sp.compute(x)
{ 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0}

Caveats

PyEtaler tries it's best to make the Python API similar to the C++ API (so ou can use the C++ documents as the Python document). There are some caveats. A few changes are made in an effort to make the Python API Pythonic. The changes are:

Tensor.toHost and Tensor.item are not templates

Since Python is a dynamically typed language. There's really no point have functions maintaining a template. Your code in C++

auto t = et::ones({4,4});
auto vec = t.toHost<int>()

becomes

t = et.ones([4,4])
vec = t.toHost()

No more brace arround tensor indices

It is quite annoning having to have extra braces when indexing. So we removed them in Python!

auto t = et::ones({4,4});
auto q = t[{2, 2}];

becomes

t = et.ones([4,4])
q = t[2, 2]

Logical operators doesn't work

Due to how Python works. It is impossible to provide the logical operators in the wrapper.

auto r = t && q;

Should have become:

r = t and q

But instead of mapping operator &&. Python will try casting et.Tensor into a bool then performing the and operation. So instead of using and. Please use the logical_and function.

r = et.logical_and(t, q)

Hacking PyEtaler

In case that you need to use C++ STL - maybe because the wrapper is doing something stupid. You can access the STL using etaler.std.

For example

>>> from etaler import std
>>> std.vector[int](10)
<cppyy.gbl.std.vector<int> object at 0x1ef112c0>

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for pyetaler, version 0.0.6
Filename, size File type Python version Upload date Hashes
Filename, size pyetaler-0.0.6.tar.gz (120.2 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page